{"id":14223,"date":"2026-10-14T07:41:18","date_gmt":"2026-10-14T07:41:18","guid":{"rendered":"https:\/\/www.voctos.com\/research\/chatgpt-appliance-brands-egypt\/"},"modified":"2026-10-14T10:27:38","modified_gmt":"2026-10-14T10:27:38","slug":"chatgpt-appliance-brands-egypt","status":"publish","type":"research","link":"https:\/\/www.voctos.com\/ar\/research\/chatgpt-appliance-brands-egypt\/","title":{"rendered":"Which appliance brands ChatGPT and Google AI Mode recommend in Egypt (380 answers studied)"},"content":{"rendered":"\r\n<div class=\"wp-block-group voctos-research\"><div class=\"wp-block-group__inner-container is-layout-flow wp-block-group-is-layout-flow\">\r\n\r\n\r\n<style>\r\n.voctos-research{--vr-gap:1.35em}\r\n.voctos-research p{margin:0 0 var(--vr-gap);line-height:1.78}\r\n.voctos-research p.vr-lede{font-size:1.15em;line-height:1.68;margin-bottom:1.5em}\r\n\/* containers whose paragraphs set their own rhythm *\/\r\n.voctos-research figcaption p,\r\n.voctos-research td p,\r\n.voctos-research th p,\r\n.voctos-research blockquote p:last-child{margin-bottom:0}\r\n.voctos-research .vr-get p{margin:0 0 1.1rem;line-height:1.6}\r\n.voctos-research h2{margin:2.6em 0 .55em;line-height:1.25}\r\n.voctos-research h3{margin:2em 0 .45em;line-height:1.3}\r\n.voctos-research h2:first-child{margin-top:0}\r\n.voctos-research ul,.voctos-research ol{margin:0 0 var(--vr-gap);padding-inline-start:1.4em}\r\n.voctos-research li{margin:.35em 0;line-height:1.7}\r\n.voctos-research strong{font-weight:650}\r\n.voctos-research code{font-size:.92em;padding:.12em .35em;border-radius:4px;\r\n  background:rgba(128,128,128,.14)}\r\n.voctos-research hr{margin:2.5em 0;border:0;border-top:1px solid rgba(128,128,128,.28)}\r\n.voctos-research figure.fig-wrap{margin:2.4rem 0}\r\n.voctos-research figure.fig-wrap svg{width:100%;height:auto;display:block}\r\n.voctos-research figure.fig-wrap figcaption{margin-top:.6rem;font-size:.88em;opacity:.75}\r\n.voctos-research table{width:100%;border-collapse:collapse;font-size:.95rem;\r\n  margin:1.6rem 0}\r\n.voctos-research table th,.voctos-research table td{\r\n  border-bottom:1px solid rgba(128,128,128,.28);padding:.6rem .7rem;text-align:left;\r\n  vertical-align:top}\r\n.voctos-research table th{font-weight:650;white-space:nowrap}\r\n.voctos-research table tr:hover td{background:rgba(128,128,128,.06)}\r\n.voctos-research[dir=\"rtl\"] table th,.voctos-research[dir=\"rtl\"] table td{text-align:right}\r\n.voctos-research .table-scroll{overflow-x:auto;margin:1.6rem 0}\r\n.voctos-research .table-scroll table{margin:0}\r\n.voctos-research blockquote{\r\n  border-inline-start:3px solid currentColor;margin:2rem 0;padding:.4rem 0 .4rem 1.1rem;\r\n  opacity:.9;font-size:1.02em}\r\n.voctos-research[dir=\"rtl\"] blockquote{padding:.4rem 1.1rem .4rem 0}\r\n<\/style>\r\n\r\n\r\n\r\n<p class=\"vr-lede wp-block-paragraph\">Fresh is the appliance brand ChatGPT recommends most in Egypt, and AI Mode&#8217;s close second. We asked logged-out ChatGPT 60 home appliance questions in Egyptian Arabic, then asked them again, in English, in Modern Standard Arabic and on Google AI Mode, and added 40 short search-style prompts on both engines. Fresh was recommended in 59 of 120 Egyptian Arabic ChatGPT answers and in 29 of 60 on AI Mode. The brands both engines name track what Egyptians search for.<\/p>\r\n\r\n\r\n\r\n<h2 class=\"wp-block-heading\">Key numbers<\/h2>\r\n\r\n\r\n\r\n<ul class=\"wp-block-list\"><li><strong>49.2%<\/strong> of Egyptian Arabic ChatGPT answers recommended Fresh (59 of 120; 95% interval 38.3% to 60.0%, resampling whole questions).<\/li><li><strong>0.91<\/strong> is the rank correlation between Egyptian search demand and ChatGPT&#8217;s recommendations across 25 brands. On Google AI Mode it was 0.93.<\/li><li><strong>24 of 40<\/strong> short search-style prompts on ChatGPT came back with sponsored product ads, against 3 of 60 full questions in Modern Standard Arabic asked just before them.<\/li><\/ul>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">Free to quote with a link to this page. Every answer and every code is in the data files linked at the end.<\/p>\r\n\r\n\r\n\r\n<h2 class=\"wp-block-heading\">How to read the numbers<\/h2>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">Three measures run through this page. We use them the same way in all our studies, so they can be compared.<\/p>\r\n\r\n\r\n\r\n<ul class=\"wp-block-list\"><li><strong>Recommendation Rate:<\/strong> the share of answers that offer a brand as a pick. Fresh&#8217;s is 31 of 60 in the first run.<\/li><li><strong>First-Pick Rate:<\/strong> the share of answers in which a brand is the first one offered.<\/li><li><strong>Stability Score:<\/strong> how alike two answers to the same question are, from 0 (no brand in common) to 1 (the same brands). It is the overlap of the two lists divided by everything either list names, averaged over questions.<\/li><\/ul>\r\n\r\n\r\n\r\n<h2 class=\"wp-block-heading\">What we found<\/h2>\r\n\r\n\r\n\r\n<ul class=\"wp-block-list\"><li><strong>52 of 60 answers in the first run recommended brands, four per answer on average.<\/strong> The other eight named only shops, asked for a budget first, or compared Egyptian and imported goods without naming anyone.<\/li><li><strong>Fresh leads every ChatGPT run and is second on AI Mode.<\/strong> It was recommended in 31 of 60 first-run answers (51.7%, 95% interval 39.3% to 63.8%), in 28 on the repeat, 27 in English, 35 in Modern Standard Arabic and 29 on Google AI Mode.<\/li><li><strong>The lists move; the leaders do not.<\/strong> Asked the same question twice, ChatGPT kept the same first pick only 24 times in 47, and gave the identical set of brands in 11 of 54 question pairs. Its Stability Score between the two Arabic runs was 0.54. Four of the top five brands were the same in both runs.<\/li><li><strong>The picks track search demand.<\/strong> Across 25 brands, the rank of Egyptian searches for &#8220;appliance plus brand&#8221; and the rank of ChatGPT recommendations correlate at 0.91 (Spearman), and between 0.79 and 0.87 in every other ChatGPT run. We added this comparison after seeing the first run&#8217;s data.<\/li><li><strong>English shrinks Egypt&#8217;s own brands; for formal Arabic we could not detect a change.<\/strong> Made-in-Egypt brands took 33.0% of recommendation slots in the second Arabic run and 24.7% in English, a gap of 8.2 points (95% interval 3.4 to 13.2). In Modern Standard Arabic they took 29.3%, a gap within the noise, though they came first less often.<\/li><li><strong>Google AI Mode agrees on the leaders but not the lists.<\/strong> It named 49 brands against ChatGPT&#8217;s 32 and put Tornado on top, recommended in 31 of 60 answers. Its Stability Score against ChatGPT&#8217;s Arabic answers was 0.40, lower than ChatGPT against itself.<\/li><li><strong>AI Mode leans on Facebook; ChatGPT never did.<\/strong> A Facebook post or group appeared among AI Mode&#8217;s source cards in 39 of 60 answers. ChatGPT cited Facebook or YouTube in none of its 280 answers.<\/li><li><strong>Short search-style prompts came with shopping ads.<\/strong> 24 of 40 prompts such as \u00ab\u0627\u0641\u0636\u0644 \u062a\u0643\u064a\u064a\u0641 \u0627\u0646\u0641\u0631\u062a\u0631 \u0641\u064a \u0645\u0635\u0631\u00bb came back with sponsored product units, against 3 of 60 full questions in Modern Standard Arabic.<\/li><li><strong>Five brands never appeared in the full Egyptian Arabic questions on ChatGPT:<\/strong> Electrolux, Whirlpool, Gree, Black+Decker and Kenwood, across 120 answers. All five are among the six least searched in Egypt. AI Mode recommended none of them for the full questions; formal Arabic recommended Whirlpool once, and the search-style prompts recommended Whirlpool and Black+Decker once each on ChatGPT and Black+Decker once on AI Mode.<\/li><li><strong>Crawler rules do not explain it.<\/strong> No robots.txt file we could read blocks OpenAI&#8217;s crawlers. Fresh, the most recommended brand, has no English Wikipedia article.<\/li><li><strong>Egyptian brands are rarely the first &#8220;best&#8221;.<\/strong> On ChatGPT they were the first pick in 2 of 20 Arabic &#8220;best brand&#8221; answers across two runs. On search-style prompts, neither engine put an Egyptian brand first for the plain &#8220;best X&#8221; prompt, but the &#8220;economy&#8221; version flipped that in 4 of 10 categories on ChatGPT and 7 of 10 on AI Mode.<\/li><li><strong>Shop sites as a group are cited about as often as brand sites.<\/strong> In the first run, shop and dealer sites were cited in 31 answers, brand sites in 27. Only 20.2% of recommendations came with the brand&#8217;s own site cited.<\/li><li><strong>Some service answers cite third-party sites.<\/strong> In 3 of the 10 spare-parts-and-service answers, ChatGPT cited sites that carry Toshiba, Sharp or Carrier in their domain name and are not the brand&#8217;s own.<\/li><\/ul>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">Sources, service sites and product pictures are reported from the first run, which is the pre-registered dataset. Each later wave was registered separately before it ran.<\/p>\r\n\r\n\r\n\r\n<h2 class=\"wp-block-heading\">Why we ran this<\/h2>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">Our last study asked ChatGPT for local businesses in Egypt and found that most recommendations were shown with map data. Appliance brands have no map card. When someone types \u00ab\u0625\u064a\u0647 \u0623\u062d\u0633\u0646 \u0645\u0627\u0631\u0643\u0629 \u062b\u0644\u0627\u062c\u0629 \u0641\u064a \u0645\u0635\u0631\u061f\u00bb, the answer is a short list of brand names, and a brand that is not on it never gets compared.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">We started from three brands most Egyptian households know, Beko, Unionaire and Fresh, and a plain question: who shows up, who does not, and why. The usual answer is a technical checklist. We wanted to see first which brands appear, and then check whether the checklist explains anything.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">After the first run we read what had already been published on AI brand recommendations, from a US appliance index to multilingual preprints. Three gaps stood out. Few studies repeat each question, none tests Egyptian Arabic against formal Arabic, and most test one engine. So the study grew in three waves, each registered before it ran.<\/p>\r\n\r\n\r\n\r\n<h2 class=\"wp-block-heading\">What we fixed before each wave<\/h2>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\"><strong>Wave 1.<\/strong> We registered a question, a brand panel and coding rules, not a hypothesis. The methodology was hashed at 22:43 UTC on 13 October 2026, two minutes before the first capture. Its SHA-256 is <code>d4dc74cf\u2026438985<\/code>, and the question bank&#8217;s is <code>e4e2d149\u202641bf2<\/code>. The panel fixed 25 brands and their Arabic spellings, but any brand named counted. Part two fixed seven checks per brand: its Egyptian site, robots.txt rules for AI crawlers, an Arabic version, Product markup, Semrush data, Wikipedia in English and Arabic, and how many answers cite it.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\"><strong>Wave 2,<\/strong> hashed at 06:08 UTC on 14 October: all 60 questions again in Arabic, plus the same 60 in English, interleaved in one shuffled order.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\"><strong>Wave 3,<\/strong> hashed at 08:09 UTC on 14 October: the 60 questions in Modern Standard Arabic on ChatGPT, and the original 60 on Google AI Mode. At D&#8217;s request, 40 search-style prompts were added on both engines; that addendum was hashed at 08:41 UTC, before any of them ran. The wave 3 plan also fixed the statistics below.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">Several things were decided after the first run&#8217;s data, and each is labelled where it appears: the search-demand comparison, a finer split of source types, the ownership checks on service sites, the market coding of product pictures, the split by question type, and moving Toshiba and Sharp from an &#8220;Elaraby&#8221; label to global brands once we learned Elaraby&#8217;s Toshiba agreements had ended. The Facebook, Elaraby and drawback-label counts were not registered before wave 3. Cautions attached to brands were coded after publication for every wave.<\/p>\r\n\r\n\r\n\r\n<h2 class=\"wp-block-heading\">Method<\/h2>\r\n\r\n\r\n\r\n<h3 class=\"wp-block-heading\">The questions<\/h3>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">Ten appliance categories crossed with six question types gives 60 questions, all in colloquial Egyptian Arabic, run in an order shuffled with a fixed seed.<\/p>\r\n\r\n\r\n\r\n<figure class=\"wp-block-table table-scroll\"><table><thead><tr><th>Dimension<\/th><th>Covered<\/th><\/tr><\/thead><tbody><tr><td>Categories<\/td><td>refrigerator, automatic washer, air conditioner, gas cooker, water heater, freezer, dishwasher, TV screen, microwave, vacuum cleaner<\/td><\/tr><tr><td>Question types<\/td><td>best brand (\u0625\u064a\u0647 \u0623\u062d\u0633\u0646 \u0645\u0627\u0631\u0643\u0629 \u062b\u0644\u0627\u062c\u0629 \u0641\u064a \u0645\u0635\u0631\u061f), best value (\u0623\u062d\u0633\u0646 \u062b\u0644\u0627\u062c\u0629 \u0633\u0639\u0631\u0647\u0627 \u0643\u0648\u064a\u0633 \u0641\u064a \u0645\u0635\u0631\u061f), lasts longest (\u0623\u0646\u0647\u064a \u0645\u0627\u0631\u0643\u0629 \u062b\u0644\u0627\u062c\u0629 \u0628\u062a\u0639\u064a\u0634 \u0623\u0643\u062a\u0631 \u0648\u0645\u0628\u062a\u0639\u0637\u0644\u0634\u061f), spare parts and service (\u0623\u0646\u0647\u064a \u0645\u0627\u0631\u0643\u0629 \u062b\u0644\u0627\u062c\u0629 \u0635\u064a\u0627\u0646\u062a\u0647\u0627 \u0648\u0642\u0637\u0639 \u063a\u064a\u0627\u0631\u0647\u0627 \u0645\u062a\u0648\u0641\u0631\u0629 \u0641\u064a \u0645\u0635\u0631\u061f), where to buy (\u0623\u0634\u062a\u0631\u064a \u062b\u0644\u0627\u062c\u0629 \u0645\u0646\u064a\u0646 \u0641\u064a \u0645\u0635\u0631 \u0628\u0623\u062d\u0633\u0646 \u0633\u0639\u0631\u061f), Egyptian or imported (\u062b\u0644\u0627\u062c\u0629 \u0645\u0635\u0631\u064a\u0629 \u0648\u0644\u0627 \u0645\u0633\u062a\u0648\u0631\u062f\u0629\u060c \u0623\u0646\u0647\u064a \u0623\u062d\u0633\u0646\u061f)<\/td><\/tr><\/tbody><\/table><\/figure>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">The same 60 questions were asked five ways:<\/p>\r\n\r\n\r\n\r\n<figure class=\"wp-block-table table-scroll\"><table><thead><tr><th>Run<\/th><th>Engine<\/th><th>Language<\/th><th>Answers<\/th><\/tr><\/thead><tbody><tr><td>First run<\/td><td>ChatGPT<\/td><td>Egyptian Arabic<\/td><td>60<\/td><\/tr><tr><td>Repeat<\/td><td>ChatGPT<\/td><td>Egyptian Arabic, word for word<\/td><td>60<\/td><\/tr><tr><td>English<\/td><td>ChatGPT<\/td><td>English translation<\/td><td>60<\/td><\/tr><tr><td>Formal Arabic<\/td><td>ChatGPT<\/td><td>Modern Standard Arabic, same product words<\/td><td>60<\/td><\/tr><tr><td>AI Mode<\/td><td>Google AI Mode<\/td><td>Egyptian Arabic, word for word<\/td><td>60<\/td><\/tr><\/tbody><\/table><\/figure>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">The formal Arabic version changes only the grammar and function words, so \u00ab\u0625\u064a\u0647 \u0623\u062d\u0633\u0646 \u0645\u0627\u0631\u0643\u0629 \u062b\u0644\u0627\u062c\u0629 \u0641\u064a \u0645\u0635\u0631\u061f\u00bb becomes \u00ab\u0645\u0627 \u0623\u0641\u0636\u0644 \u0645\u0627\u0631\u0643\u0629 \u062b\u0644\u0627\u062c\u0629 \u0641\u064a \u0645\u0635\u0631\u061f\u00bb. The product words stay the same, so the contrast is register, not vocabulary. The 40 search-style prompts cover each category four ways: the plain \u00ab\u0627\u0641\u0636\u0644 \u062a\u0643\u064a\u064a\u0641 \u0641\u064a \u0645\u0635\u0631\u00bb, an economy version, and two attributes such as inverter, no-frost or a size, typed without hamza the way people type into a search box.<\/p>\r\n\r\n\r\n\r\n<h3 class=\"wp-block-heading\">The instrument<\/h3>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">Every ChatGPT question went to chatgpt.com, logged out, in a new chat, from an automated browser on a machine in Egypt. Logged out means no account memory; ChatGPT still infers a location from the connection. Before each send the composer was cleared and its contents checked against the question. The full answer, every link, every source chip, the product picture strip and any sponsored unit were saved.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">Every AI Mode prompt went to google.com in AI Mode, signed out of Google, as a fresh search from the same machine. Its source links are Google redirects, so sources are read from the domains and site names on the page.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">The captures ran on 13 and 14 October 2026. The first 60 ran between 22:45 and 23:14 UTC on 13 October; the second wave between 06:09 and 07:31 UTC; the third between 08:09 and 09:51 UTC. During the second wave logged-out ChatGPT showed its message limit after 75 questions; we waited and did not work around it. Google showed no verification challenge. Every exported file matches a hash computed in the browser before export.<\/p>\r\n\r\n\r\n\r\n<h3 class=\"wp-block-heading\">Coding<\/h3>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">Every brand named in an answer is coded once. It is <strong>recommended<\/strong> when the answer offers it as a pick: a heading, a list item, a product to buy, or wording such as &#8220;start by comparing X and Y&#8221;. It is <strong>mentioned<\/strong> when it is only a comparison point, a price example inside a shop recommendation, or one of &#8220;brands like&#8221;. The first recommended brand is the first pick. Options inside the closing question form are not counted.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">Brands created in Egypt are grouped as made in Egypt: Fresh, Tornado, Kiriazi, Unionaire, White Whale, Universal and Olympic Electric, plus the state Military Factories that AI Mode named. Olympic Electric&#8217;s maker, Olympic Group, has been owned by Electrolux since 2011 (<a href=\"https:\/\/www.electroluxgroup.com\/en\/electrolux-has-completed-the-acquisition-of-olympic-group-11655\" target=\"_blank\" rel=\"noopener\">Electrolux<\/a>), but the brand itself is Egyptian. Sharp, made in Egypt by Elaraby, and Toshiba count as global brands.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">Each wave after the first was coded by coders who worked from the same written rules without sight of earlier codes. Several brand names are also ordinary words: \u00ab\u0645\u064a\u062f\u064a\u0627\u00bb, \u00ab\u0634\u0627\u0631\u0628\u00bb, \u00ab\u0641\u0631\u064a\u0634\u00bb, \u00ab\u062a\u0648\u0631\u0646\u064a\u062f\u0648\u00bb. We printed every line where one of nine such names appeared in an answer that had not coded that brand. That was 4 of 741 matching lines, and all four were options in ChatGPT&#8217;s closing question, which the rules exclude.<\/p>\r\n\r\n\r\n\r\n<h3 class=\"wp-block-heading\">The audit and the demand measure<\/h3>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">For part two we found each brand&#8217;s Egyptian site from its own channels, read its robots.txt, checked for an Arabic version and looked for Product markup on one product page. We pulled Semrush&#8217;s Egypt figures for each site and checked Wikipedia.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">Demand was added after the data. For each brand we summed Semrush&#8217;s monthly Egyptian search volume for the ten appliance words plus the brand name, in Arabic and Latin spelling, 580 phrases in all. A phrase with no Semrush data counts as zero.<\/p>\r\n\r\n\r\n\r\n<h3 class=\"wp-block-heading\">Statistics<\/h3>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">Answers to the same question are not independent. A rate pooled over runs counts each question two or more times, so its interval resamples whole questions with all their runs, 5,000 times. That widens Fresh&#8217;s pooled Arabic interval to 38.3% to 60.0%, against 40.4% to 58.0% if every answer is treated as separate. Within one run, key shares also get a bootstrap over answers: 23.6% to 36.4% for the first-run made-in-Egypt share, against 24.0% to 36.3% from Wilson.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">Rank correlations are descriptive. With 25 brands we report them and run no tests, but we resampled the brands 5,000 times: the first-run correlation of 0.91 has a bootstrap range of 0.75 to 0.98.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">To see what drives the variation, we split it the way reliability studies do: for every brand, question and run, was the brand recommended or not, and how much of the spread comes from the brand, the question, the run, and their combinations. The method is a generalizability study on a fully crossed design, with variance components from expected mean squares.<\/p>\r\n\r\n\r\n\r\n<h2 class=\"wp-block-heading\">Which brands does ChatGPT recommend for appliances in Egypt?<\/h2>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">In the first run we counted 208 recommendations across 52 answers, and 32 brands named in total. Fresh was recommended in 31 answers, Toshiba and Beko in 21, LG in 18, Sharp in 14 and Tornado in 12. The top three brands took 35.1% of all recommendation slots, and the top five took 50.5%.<\/p>\r\n\r\n\r\n\r\n<figure class=\"wp-block-table table-scroll\"><table><thead><tr><th>Brand<\/th><th>Made in<\/th><th>Recommended in (of 60)<\/th><th>Named in<\/th><th>First pick<\/th><th>Categories<\/th><\/tr><\/thead><tbody><tr><td>Fresh<\/td><td>Egypt<\/td><td>31<\/td><td>34<\/td><td>14<\/td><td>9<\/td><\/tr><tr><td>Toshiba<\/td><td>global<\/td><td>21<\/td><td>23<\/td><td>9<\/td><td>7<\/td><\/tr><tr><td>Beko<\/td><td>global<\/td><td>21<\/td><td>21<\/td><td>2<\/td><td>7<\/td><\/tr><tr><td>LG<\/td><td>global<\/td><td>18<\/td><td>20<\/td><td>2<\/td><td>6<\/td><\/tr><tr><td>Sharp<\/td><td>global<\/td><td>14<\/td><td>16<\/td><td>5<\/td><td>5<\/td><\/tr><tr><td>Tornado<\/td><td>Egypt<\/td><td>12<\/td><td>13<\/td><td>1<\/td><td>8<\/td><\/tr><tr><td>Samsung<\/td><td>global<\/td><td>11<\/td><td>15<\/td><td>3<\/td><td>5<\/td><\/tr><tr><td>Bosch<\/td><td>global<\/td><td>11<\/td><td>11<\/td><td>4<\/td><td>4<\/td><\/tr><tr><td>Zanussi<\/td><td>global<\/td><td>9<\/td><td>11<\/td><td>2<\/td><td>4<\/td><\/tr><tr><td>Midea<\/td><td>global<\/td><td>7<\/td><td>8<\/td><td>1<\/td><td>4<\/td><\/tr><tr><td>Kiriazi<\/td><td>Egypt<\/td><td>7<\/td><td>7<\/td><td>2<\/td><td>3<\/td><\/tr><\/tbody><\/table><\/figure>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">The leader changes by category. Beko was recommended in all 6 dishwasher answers that named brands, and Fresh in all 6 freezer answers. Toshiba led refrigerators with 5, Samsung led TV screens with 4, and Philips led vacuum cleaners with 4.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">Two of the six TV questions came back about computer monitors, because ChatGPT read \u00ab\u0634\u0627\u0634\u0629\u00bb as a monitor and recommended Gigabyte and MSI instead of television brands. We kept those answers as captured, because a buyer typing the same words gets the same reading. The repeat run did it again in 2 of 6, and the formal Arabic run in 2 of 6; the English &#8220;TV&#8221; questions and AI Mode never did.<\/p>\r\n\r\n\r\n\r\n<h2 class=\"wp-block-heading\">Does ChatGPT give the same answer twice?<\/h2>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">Not in detail, but the leaders hold. We asked the same 60 Arabic questions a second time and compared each answer with its twin. The first pick was the same in 24 of the 47 pairs where both answers named one. The two answers named exactly the same brands in 11 of the 54 pairs that recommended anything, and the Stability Score was 0.54.<\/p>\r\n\r\n\r\n\r\n<figure class=\"wp-block-table table-scroll\"><table><thead><tr><th>Brand<\/th><th>First run (of 60)<\/th><th>Second run (of 60)<\/th><th>Both runs<\/th><\/tr><\/thead><tbody><tr><td>Fresh<\/td><td>31<\/td><td>28<\/td><td>21<\/td><\/tr><tr><td>Toshiba<\/td><td>21<\/td><td>15<\/td><td>14<\/td><\/tr><tr><td>Beko<\/td><td>21<\/td><td>12<\/td><td>11<\/td><\/tr><tr><td>LG<\/td><td>18<\/td><td>16<\/td><td>14<\/td><\/tr><tr><td>Sharp<\/td><td>14<\/td><td>14<\/td><td>11<\/td><\/tr><tr><td>Tornado<\/td><td>12<\/td><td>15<\/td><td>7<\/td><\/tr><tr><td>Samsung<\/td><td>11<\/td><td>13<\/td><td>8<\/td><\/tr><\/tbody><\/table><\/figure>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">Brand by brand, the counts stayed close for most of the leaders, and four of the top five were the same in both runs. Beko was the exception: it fell from 21 answers to 12, and Tornado took its place in the top five. Pooled over both runs, Fresh appeared in 59 of 120 Arabic answers, Toshiba in 36 and LG in 34.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">The direction matches a larger study. SparkToro and Gumshoe ran a dozen prompts about 3,000 times in total and found under a 1% chance of the same list twice, while the leading brands kept appearing (<a href=\"https:\/\/searchengineland.com\/ai-recommendation-lists-rarely-repeat-study-468076\" target=\"_blank\" rel=\"noopener\">Search Engine Land<\/a>). Our identical-list rate, 11 of 54, is far higher than theirs, likely because our questions are narrower and our lists shorter. <strong>Read any single ChatGPT answer as a draw from a distribution. Read the brand counts across many answers as the signal.<\/strong><\/p>\r\n\r\n\r\n\r\n<h2 class=\"wp-block-heading\">Does the language of the question change the picks?<\/h2>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">English does; formal Arabic changes less than we could measure. The table sets each version of the 60 questions against the second Egyptian Arabic run, with the two Arabic runs against each other as the baseline for ordinary run-to-run change.<\/p>\r\n\r\n\r\n\r\n<figure class=\"wp-block-table table-scroll\"><table><thead><tr><th>Compared with the repeat run<\/th><th>Stability Score<\/th><th>Same first pick<\/th><th>Made-in-Egypt share of slots<\/th><\/tr><\/thead><tbody><tr><td>First run, same words<\/td><td>0.54<\/td><td>24 of 47<\/td><td>29.8% against 33.0%<\/td><\/tr><tr><td>Modern Standard Arabic<\/td><td>0.52<\/td><td>26 of 48<\/td><td>29.3% against 33.0%<\/td><\/tr><tr><td>English<\/td><td>0.40<\/td><td>23 of 48<\/td><td>24.7% against 33.0%<\/td><\/tr><\/tbody><\/table><\/figure>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">Formal Arabic moved the lists about as much as asking again in the same words. Its made-in-Egypt share sat 3.7 points below the repeat run, with an interval from -1.2 to 8.6 that includes no change. First picks lean the other way, though: an Egyptian brand came first in 13 of 53 formal Arabic answers, against 18 of 49 in the repeat and 16 of 54 in English. Fresh did well in it: recommended in 35 of 60 answers.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">English is different. Its answers overlapped less with the Arabic ones, and the made-in-Egypt share fell by 8.2 points, an interval of 3.4 to 13.2 that excludes zero. Bosch was recommended in 18 English answers against 9 Arabic ones, while Sharp fell from 14 to 7, Tornado from 15 to 8 and Kiriazi from 8 to 3. Universal, recommended in 5 Arabic answers, did not appear in English at all. Paired question by question, an Egyptian brand was recommended in the Arabic answer but not the English one 11 times, and the reverse 4 times.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">Unionaire went the other way, from 3 Arabic answers to 9 English ones, and English answers named more brands overall: 41 against 32. Counting only domains we had already classified in the first run, they also cited brands&#8217; own sites more often, in 40 answers against 32.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">The direction fits a 2026 preprint that ran identical prompts in several languages on logged-out ChatGPT and found that the language of the question, more than the user&#8217;s location, decides whether local suppliers appear (<a href=\"https:\/\/arxiv.org\/abs\/2608.30052\" target=\"_blank\" rel=\"noopener\">arXiv<\/a>). <strong>For an Egyptian brand, Arabic answers are home ground. English answers are where global brands gain.<\/strong><\/p>\r\n\r\n\r\n\r\n<h2 class=\"wp-block-heading\">Does Google AI Mode recommend the same brands?<\/h2>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">The same leaders, a different list. We asked AI Mode the original 60 Egyptian Arabic questions word for word. It recommended brands in 49 answers and named 49 brands in all, against 32 in ChatGPT&#8217;s first run.<\/p>\r\n\r\n\r\n\r\n<figure class=\"wp-block-table table-scroll\"><table><thead><tr><th>Brand<\/th><th>ChatGPT, first run (of 60)<\/th><th>ChatGPT, repeat (of 60)<\/th><th>Google AI Mode (of 60)<\/th><\/tr><\/thead><tbody><tr><td>Tornado<\/td><td>12<\/td><td>15<\/td><td>31<\/td><\/tr><tr><td>Fresh<\/td><td>31<\/td><td>28<\/td><td>29<\/td><\/tr><tr><td>LG<\/td><td>18<\/td><td>16<\/td><td>24<\/td><\/tr><tr><td>Sharp<\/td><td>14<\/td><td>14<\/td><td>21<\/td><\/tr><tr><td>Toshiba<\/td><td>21<\/td><td>15<\/td><td>20<\/td><\/tr><tr><td>Beko<\/td><td>21<\/td><td>12<\/td><td>18<\/td><\/tr><tr><td>Samsung<\/td><td>11<\/td><td>13<\/td><td>16<\/td><\/tr><\/tbody><\/table><\/figure>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">AI Mode&#8217;s picks track search demand about as closely as ChatGPT&#8217;s, at 0.93 across the 25 brands. Made-in-Egypt brands took 29.1% of its slots, close to ChatGPT. The difference is order. AI Mode put an Egyptian brand first in 9 of 49 answers that recommended brands, against 18 of 52 in ChatGPT&#8217;s first run.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">The answers themselves differ more than ChatGPT&#8217;s do between runs. Against ChatGPT&#8217;s repeat run, AI Mode&#8217;s Stability Score was 0.40 and the first pick matched in 15 of 46 questions. None of the five brands ChatGPT never named was recommended by AI Mode either; Electrolux, Black+Decker and Kenwood were mentioned, never offered.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">AI Mode also argues more. 14 of its 60 answers carried an explicit drawbacks label, such as \u00ab\u0627\u0644\u0639\u064a\u0648\u0628\u00bb, under recommended brands, against 3 of 180 ChatGPT answers in the repeat, English and formal Arabic runs.<\/p>\r\n\r\n\r\n\r\n<figure class=\"fig-wrap\" data-fig=\"fig-runs.svg\"><svg class=\"fig\" xmlns=\"http:\/\/www.w3.org\/2000\/svg\" viewBox=\"0 0 900 724\" role=\"img\" aria-label=\"Five ways to ask: Fresh holds, English lifts Bosch, AI Mode lifts Tornado. Answers (of 60) recommending each brand, for the same 60 questions asked five ways.\">\r\n  <style>\r\n    .fig { --ink:#12203a; --mut:#5a6b86; --grid:#dfe5ee; --bg:#ffffff; --glob:#1f5f8b; --home:#b4532a; --pale:#d6dde8; }\r\n    @media (prefers-color-scheme: dark) {\r\n      .fig { --ink:#e8edf5; --mut:#9dabc2; --grid:#2b3648; --bg:#0f1520; --glob:#6fb2dd; --home:#e0855c; --pale:#38445a; }\r\n    }\r\n    .fig text { font-family: ui-sans-serif, system-ui, \"Segoe UI\", Arial, sans-serif; fill: var(--ink); }\r\n    .fig .mut { fill: var(--mut); }\r\n    .fig .grid { stroke: var(--grid); stroke-width: 1; }\r\n    .fig .ttl { font-weight: 700; }\r\n  <\/style>\r\n<rect width=\"900\" height=\"724\" fill=\"var(--bg)\"\/>\r\n<text x=\"24\" y=\"32\" class=\"ttl\" font-size=\"19\">Five ways to ask: Fresh holds, English lifts Bosch, AI Mode lifts Tornado<\/text>\r\n<text x=\"24\" y=\"54\" class=\"mut\" font-size=\"13\">Answers (of 60) recommending each brand, for the same 60 questions asked five ways.<\/text>\r\n\r\n<rect x=\"24\" y=\"66\" width=\"12\" height=\"12\" fill=\"var(--home)\" opacity=\"1.0\"\/><text x=\"42\" y=\"76\" font-size=\"12\">ChatGPT, Arabic run 1<\/text>\r\n<rect x=\"224\" y=\"66\" width=\"12\" height=\"12\" fill=\"var(--home)\" opacity=\"0.55\"\/><text x=\"242\" y=\"76\" font-size=\"12\">ChatGPT, Arabic run 2<\/text>\r\n<rect x=\"424\" y=\"66\" width=\"12\" height=\"12\" fill=\"var(--home)\" opacity=\"0.3\"\/><text x=\"442\" y=\"76\" font-size=\"12\">ChatGPT, formal Arabic<\/text>\r\n<rect x=\"24\" y=\"86\" width=\"12\" height=\"12\" fill=\"var(--glob)\" opacity=\"1.0\"\/><text x=\"42\" y=\"96\" font-size=\"12\">ChatGPT, English<\/text>\r\n<rect x=\"224\" y=\"86\" width=\"12\" height=\"12\" fill=\"var(--ink)\" opacity=\"0.85\"\/><text x=\"242\" y=\"96\" font-size=\"12\">Google AI Mode<\/text>\r\n<text x=\"140\" y=\"130\" font-size=\"13\" text-anchor=\"end\">Fresh<\/text>\r\n<rect x=\"150\" y=\"114.0\" width=\"576.9\" height=\"6.6\" fill=\"var(--home)\" opacity=\"1.0\"\/>\r\n<text x=\"731.9\" y=\"119.6\" font-size=\"9\" class=\"mut\">31<\/text>\r\n<rect x=\"150\" y=\"121.6\" width=\"521.1\" height=\"6.6\" fill=\"var(--home)\" opacity=\"0.55\"\/>\r\n<text x=\"676.1\" y=\"127.2\" font-size=\"9\" class=\"mut\">28<\/text>\r\n<rect x=\"150\" y=\"129.2\" width=\"651.4\" height=\"6.6\" fill=\"var(--home)\" opacity=\"0.3\"\/>\r\n<text x=\"806.4\" y=\"134.8\" font-size=\"9\" class=\"mut\">35<\/text>\r\n<rect x=\"150\" y=\"136.8\" width=\"502.5\" height=\"6.6\" fill=\"var(--glob)\" opacity=\"1.0\"\/>\r\n<text x=\"657.5\" y=\"142.4\" font-size=\"9\" class=\"mut\">27<\/text>\r\n<rect x=\"150\" y=\"144.4\" width=\"539.7\" height=\"6.6\" fill=\"var(--ink)\" opacity=\"0.85\"\/>\r\n<text x=\"694.7\" y=\"150.0\" font-size=\"9\" class=\"mut\">29<\/text>\r\n<text x=\"140\" y=\"176\" font-size=\"13\" text-anchor=\"end\">Tornado<\/text>\r\n<rect x=\"150\" y=\"160.0\" width=\"223.3\" height=\"6.6\" fill=\"var(--home)\" opacity=\"1.0\"\/>\r\n<text x=\"378.3\" y=\"165.6\" font-size=\"9\" class=\"mut\">12<\/text>\r\n<rect x=\"150\" y=\"167.6\" width=\"279.2\" height=\"6.6\" fill=\"var(--home)\" opacity=\"0.55\"\/>\r\n<text x=\"434.2\" y=\"173.2\" font-size=\"9\" class=\"mut\">15<\/text>\r\n<rect x=\"150\" y=\"175.2\" width=\"167.5\" height=\"6.6\" fill=\"var(--home)\" opacity=\"0.3\"\/>\r\n<text x=\"322.5\" y=\"180.8\" font-size=\"9\" class=\"mut\">9<\/text>\r\n<rect x=\"150\" y=\"182.8\" width=\"148.9\" height=\"6.6\" fill=\"var(--glob)\" opacity=\"1.0\"\/>\r\n<text x=\"303.9\" y=\"188.4\" font-size=\"9\" class=\"mut\">8<\/text>\r\n<rect x=\"150\" y=\"190.4\" width=\"576.9\" height=\"6.6\" fill=\"var(--ink)\" opacity=\"0.85\"\/>\r\n<text x=\"731.9\" y=\"196.0\" font-size=\"9\" class=\"mut\">31<\/text>\r\n<text x=\"140\" y=\"222\" font-size=\"13\" text-anchor=\"end\">LG<\/text>\r\n<rect x=\"150\" y=\"206.0\" width=\"335.0\" height=\"6.6\" fill=\"var(--home)\" opacity=\"1.0\"\/>\r\n<text x=\"490.0\" y=\"211.6\" font-size=\"9\" class=\"mut\">18<\/text>\r\n<rect x=\"150\" y=\"213.6\" width=\"297.8\" height=\"6.6\" fill=\"var(--home)\" opacity=\"0.55\"\/>\r\n<text x=\"452.8\" y=\"219.2\" font-size=\"9\" class=\"mut\">16<\/text>\r\n<rect x=\"150\" y=\"221.2\" width=\"353.6\" height=\"6.6\" fill=\"var(--home)\" opacity=\"0.3\"\/>\r\n<text x=\"508.6\" y=\"226.8\" font-size=\"9\" class=\"mut\">19<\/text>\r\n<rect x=\"150\" y=\"228.8\" width=\"353.6\" height=\"6.6\" fill=\"var(--glob)\" opacity=\"1.0\"\/>\r\n<text x=\"508.6\" y=\"234.4\" font-size=\"9\" class=\"mut\">19<\/text>\r\n<rect x=\"150\" y=\"236.4\" width=\"446.7\" height=\"6.6\" fill=\"var(--ink)\" opacity=\"0.85\"\/>\r\n<text x=\"601.7\" y=\"242.0\" font-size=\"9\" class=\"mut\">24<\/text>\r\n<text x=\"140\" y=\"268\" font-size=\"13\" text-anchor=\"end\">Toshiba<\/text>\r\n<rect x=\"150\" y=\"252.0\" width=\"390.8\" height=\"6.6\" fill=\"var(--home)\" opacity=\"1.0\"\/>\r\n<text x=\"545.8\" y=\"257.6\" font-size=\"9\" class=\"mut\">21<\/text>\r\n<rect x=\"150\" y=\"259.6\" width=\"279.2\" height=\"6.6\" fill=\"var(--home)\" opacity=\"0.55\"\/>\r\n<text x=\"434.2\" y=\"265.2\" font-size=\"9\" class=\"mut\">15<\/text>\r\n<rect x=\"150\" y=\"267.2\" width=\"335.0\" height=\"6.6\" fill=\"var(--home)\" opacity=\"0.3\"\/>\r\n<text x=\"490.0\" y=\"272.8\" font-size=\"9\" class=\"mut\">18<\/text>\r\n<rect x=\"150\" y=\"274.8\" width=\"260.6\" height=\"6.6\" fill=\"var(--glob)\" opacity=\"1.0\"\/>\r\n<text x=\"415.6\" y=\"280.4\" font-size=\"9\" class=\"mut\">14<\/text>\r\n<rect x=\"150\" y=\"282.4\" width=\"372.2\" height=\"6.6\" fill=\"var(--ink)\" opacity=\"0.85\"\/>\r\n<text x=\"527.2\" y=\"288.0\" font-size=\"9\" class=\"mut\">20<\/text>\r\n<text x=\"140\" y=\"314\" font-size=\"13\" text-anchor=\"end\">Sharp<\/text>\r\n<rect x=\"150\" y=\"298.0\" width=\"260.6\" height=\"6.6\" fill=\"var(--home)\" opacity=\"1.0\"\/>\r\n<text x=\"415.6\" y=\"303.6\" font-size=\"9\" class=\"mut\">14<\/text>\r\n<rect x=\"150\" y=\"305.6\" width=\"260.6\" height=\"6.6\" fill=\"var(--home)\" opacity=\"0.55\"\/>\r\n<text x=\"415.6\" y=\"311.2\" font-size=\"9\" class=\"mut\">14<\/text>\r\n<rect x=\"150\" y=\"313.2\" width=\"297.8\" height=\"6.6\" fill=\"var(--home)\" opacity=\"0.3\"\/>\r\n<text x=\"452.8\" y=\"318.8\" font-size=\"9\" class=\"mut\">16<\/text>\r\n<rect x=\"150\" y=\"320.8\" width=\"130.3\" height=\"6.6\" fill=\"var(--glob)\" opacity=\"1.0\"\/>\r\n<text x=\"285.3\" y=\"326.4\" font-size=\"9\" class=\"mut\">7<\/text>\r\n<rect x=\"150\" y=\"328.4\" width=\"390.8\" height=\"6.6\" fill=\"var(--ink)\" opacity=\"0.85\"\/>\r\n<text x=\"545.8\" y=\"334.0\" font-size=\"9\" class=\"mut\">21<\/text>\r\n<text x=\"140\" y=\"360\" font-size=\"13\" text-anchor=\"end\">Beko<\/text>\r\n<rect x=\"150\" y=\"344.0\" width=\"390.8\" height=\"6.6\" fill=\"var(--home)\" opacity=\"1.0\"\/>\r\n<text x=\"545.8\" y=\"349.6\" font-size=\"9\" class=\"mut\">21<\/text>\r\n<rect x=\"150\" y=\"351.6\" width=\"223.3\" height=\"6.6\" fill=\"var(--home)\" opacity=\"0.55\"\/>\r\n<text x=\"378.3\" y=\"357.2\" font-size=\"9\" class=\"mut\">12<\/text>\r\n<rect x=\"150\" y=\"359.2\" width=\"335.0\" height=\"6.6\" fill=\"var(--home)\" opacity=\"0.3\"\/>\r\n<text x=\"490.0\" y=\"364.8\" font-size=\"9\" class=\"mut\">18<\/text>\r\n<rect x=\"150\" y=\"366.8\" width=\"204.7\" height=\"6.6\" fill=\"var(--glob)\" opacity=\"1.0\"\/>\r\n<text x=\"359.7\" y=\"372.4\" font-size=\"9\" class=\"mut\">11<\/text>\r\n<rect x=\"150\" y=\"374.4\" width=\"335.0\" height=\"6.6\" fill=\"var(--ink)\" opacity=\"0.85\"\/>\r\n<text x=\"490.0\" y=\"380.0\" font-size=\"9\" class=\"mut\">18<\/text>\r\n<text x=\"140\" y=\"406\" font-size=\"13\" text-anchor=\"end\">Samsung<\/text>\r\n<rect x=\"150\" y=\"390.0\" width=\"204.7\" height=\"6.6\" fill=\"var(--home)\" opacity=\"1.0\"\/>\r\n<text x=\"359.7\" y=\"395.6\" font-size=\"9\" class=\"mut\">11<\/text>\r\n<rect x=\"150\" y=\"397.6\" width=\"241.9\" height=\"6.6\" fill=\"var(--home)\" opacity=\"0.55\"\/>\r\n<text x=\"396.9\" y=\"403.2\" font-size=\"9\" class=\"mut\">13<\/text>\r\n<rect x=\"150\" y=\"405.2\" width=\"223.3\" height=\"6.6\" fill=\"var(--home)\" opacity=\"0.3\"\/>\r\n<text x=\"378.3\" y=\"410.8\" font-size=\"9\" class=\"mut\">12<\/text>\r\n<rect x=\"150\" y=\"412.8\" width=\"204.7\" height=\"6.6\" fill=\"var(--glob)\" opacity=\"1.0\"\/>\r\n<text x=\"359.7\" y=\"418.4\" font-size=\"9\" class=\"mut\">11<\/text>\r\n<rect x=\"150\" y=\"420.4\" width=\"297.8\" height=\"6.6\" fill=\"var(--ink)\" opacity=\"0.85\"\/>\r\n<text x=\"452.8\" y=\"426.0\" font-size=\"9\" class=\"mut\">16<\/text>\r\n<text x=\"140\" y=\"452\" font-size=\"13\" text-anchor=\"end\">Bosch<\/text>\r\n<rect x=\"150\" y=\"436.0\" width=\"204.7\" height=\"6.6\" fill=\"var(--home)\" opacity=\"1.0\"\/>\r\n<text x=\"359.7\" y=\"441.6\" font-size=\"9\" class=\"mut\">11<\/text>\r\n<rect x=\"150\" y=\"443.6\" width=\"167.5\" height=\"6.6\" fill=\"var(--home)\" opacity=\"0.55\"\/>\r\n<text x=\"322.5\" y=\"449.2\" font-size=\"9\" class=\"mut\">9<\/text>\r\n<rect x=\"150\" y=\"451.2\" width=\"148.9\" height=\"6.6\" fill=\"var(--home)\" opacity=\"0.3\"\/>\r\n<text x=\"303.9\" y=\"456.8\" font-size=\"9\" class=\"mut\">8<\/text>\r\n<rect x=\"150\" y=\"458.8\" width=\"335.0\" height=\"6.6\" fill=\"var(--glob)\" opacity=\"1.0\"\/>\r\n<text x=\"490.0\" y=\"464.4\" font-size=\"9\" class=\"mut\">18<\/text>\r\n<rect x=\"150\" y=\"466.4\" width=\"223.3\" height=\"6.6\" fill=\"var(--ink)\" opacity=\"0.85\"\/>\r\n<text x=\"378.3\" y=\"472.0\" font-size=\"9\" class=\"mut\">12<\/text>\r\n<text x=\"140\" y=\"498\" font-size=\"13\" text-anchor=\"end\">Zanussi<\/text>\r\n<rect x=\"150\" y=\"482.0\" width=\"167.5\" height=\"6.6\" fill=\"var(--home)\" opacity=\"1.0\"\/>\r\n<text x=\"322.5\" y=\"487.6\" font-size=\"9\" class=\"mut\">9<\/text>\r\n<rect x=\"150\" y=\"489.6\" width=\"204.7\" height=\"6.6\" fill=\"var(--home)\" opacity=\"0.55\"\/>\r\n<text x=\"359.7\" y=\"495.2\" font-size=\"9\" class=\"mut\">11<\/text>\r\n<rect x=\"150\" y=\"497.2\" width=\"223.3\" height=\"6.6\" fill=\"var(--home)\" opacity=\"0.3\"\/>\r\n<text x=\"378.3\" y=\"502.8\" font-size=\"9\" class=\"mut\">12<\/text>\r\n<rect x=\"150\" y=\"504.8\" width=\"186.1\" height=\"6.6\" fill=\"var(--glob)\" opacity=\"1.0\"\/>\r\n<text x=\"341.1\" y=\"510.4\" font-size=\"9\" class=\"mut\">10<\/text>\r\n<rect x=\"150\" y=\"512.4\" width=\"223.3\" height=\"6.6\" fill=\"var(--ink)\" opacity=\"0.85\"\/>\r\n<text x=\"378.3\" y=\"518.0\" font-size=\"9\" class=\"mut\">12<\/text>\r\n<text x=\"140\" y=\"544\" font-size=\"13\" text-anchor=\"end\">Midea<\/text>\r\n<rect x=\"150\" y=\"528.0\" width=\"130.3\" height=\"6.6\" fill=\"var(--home)\" opacity=\"1.0\"\/>\r\n<text x=\"285.3\" y=\"533.6\" font-size=\"9\" class=\"mut\">7<\/text>\r\n<rect x=\"150\" y=\"535.6\" width=\"186.1\" height=\"6.6\" fill=\"var(--home)\" opacity=\"0.55\"\/>\r\n<text x=\"341.1\" y=\"541.2\" font-size=\"9\" class=\"mut\">10<\/text>\r\n<rect x=\"150\" y=\"543.2\" width=\"167.5\" height=\"6.6\" fill=\"var(--home)\" opacity=\"0.3\"\/>\r\n<text x=\"322.5\" y=\"548.8\" font-size=\"9\" class=\"mut\">9<\/text>\r\n<rect x=\"150\" y=\"550.8\" width=\"167.5\" height=\"6.6\" fill=\"var(--glob)\" opacity=\"1.0\"\/>\r\n<text x=\"322.5\" y=\"556.4\" font-size=\"9\" class=\"mut\">9<\/text>\r\n<rect x=\"150\" y=\"558.4\" width=\"130.3\" height=\"6.6\" fill=\"var(--ink)\" opacity=\"0.85\"\/>\r\n<text x=\"285.3\" y=\"564.0\" font-size=\"9\" class=\"mut\">7<\/text>\r\n<text x=\"140\" y=\"590\" font-size=\"13\" text-anchor=\"end\">Unionaire<\/text>\r\n<rect x=\"150\" y=\"574.0\" width=\"55.8\" height=\"6.6\" fill=\"var(--home)\" opacity=\"1.0\"\/>\r\n<text x=\"210.8\" y=\"579.6\" font-size=\"9\" class=\"mut\">3<\/text>\r\n<rect x=\"150\" y=\"581.6\" width=\"55.8\" height=\"6.6\" fill=\"var(--home)\" opacity=\"0.55\"\/>\r\n<text x=\"210.8\" y=\"587.2\" font-size=\"9\" class=\"mut\">3<\/text>\r\n<rect x=\"150\" y=\"589.2\" width=\"93.1\" height=\"6.6\" fill=\"var(--home)\" opacity=\"0.3\"\/>\r\n<text x=\"248.1\" y=\"594.8\" font-size=\"9\" class=\"mut\">5<\/text>\r\n<rect x=\"150\" y=\"596.8\" width=\"167.5\" height=\"6.6\" fill=\"var(--glob)\" opacity=\"1.0\"\/>\r\n<text x=\"322.5\" y=\"602.4\" font-size=\"9\" class=\"mut\">9<\/text>\r\n<rect x=\"150\" y=\"604.4\" width=\"37.2\" height=\"6.6\" fill=\"var(--ink)\" opacity=\"0.85\"\/>\r\n<text x=\"192.2\" y=\"610.0\" font-size=\"9\" class=\"mut\">2<\/text>\r\n<text x=\"140\" y=\"636\" font-size=\"13\" text-anchor=\"end\">Kiriazi<\/text>\r\n<rect x=\"150\" y=\"620.0\" width=\"130.3\" height=\"6.6\" fill=\"var(--home)\" opacity=\"1.0\"\/>\r\n<text x=\"285.3\" y=\"625.6\" font-size=\"9\" class=\"mut\">7<\/text>\r\n<rect x=\"150\" y=\"627.6\" width=\"148.9\" height=\"6.6\" fill=\"var(--home)\" opacity=\"0.55\"\/>\r\n<text x=\"303.9\" y=\"633.2\" font-size=\"9\" class=\"mut\">8<\/text>\r\n<rect x=\"150\" y=\"635.2\" width=\"130.3\" height=\"6.6\" fill=\"var(--home)\" opacity=\"0.3\"\/>\r\n<text x=\"285.3\" y=\"640.8\" font-size=\"9\" class=\"mut\">7<\/text>\r\n<rect x=\"150\" y=\"642.8\" width=\"55.8\" height=\"6.6\" fill=\"var(--glob)\" opacity=\"1.0\"\/>\r\n<text x=\"210.8\" y=\"648.4\" font-size=\"9\" class=\"mut\">3<\/text>\r\n<rect x=\"150\" y=\"650.4\" width=\"74.4\" height=\"6.6\" fill=\"var(--ink)\" opacity=\"0.85\"\/>\r\n<text x=\"229.4\" y=\"656.0\" font-size=\"9\" class=\"mut\">4<\/text>\r\n<text x=\"24\" y=\"710\" font-size=\"11\" class=\"mut\">Source: VOCTOS, 240 logged-out ChatGPT answers and 60 Google AI Mode answers captured in Egypt, 13 to 14 October 2026.<\/text>\r\n<\/svg><figcaption>Five ways to ask: Fresh holds, English lifts Bosch, AI Mode lifts Tornado. Answers (of 60) recommending each brand, for the same 60 questions asked five ways.<\/figcaption><\/figure>\r\n\r\n\r\n\r\n<h2 class=\"wp-block-heading\">Which sources does each engine use?<\/h2>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">ChatGPT cites shops and brands. In the first run we found at least one cited source in 58 of 60 answers, across 64 domains. Shop and dealer sites appeared in 31 answers and brand sites in 27. B.TECH&#8217;s and Jumia&#8217;s sites were each cited in 12 answers, Carrefour Egypt&#8217;s in 11, and one Egyptian review site, Mizaan Home, in 8.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">Per recommendation, the brand&#8217;s own site is rarer still. Of 208 recommendations, 42 came with the brand&#8217;s own site cited somewhere in the answer (20.2%, 95% interval 15.3% to 26.2%). LG managed 8 of its 18, Bosch 6 of 11 and Fresh 6 of 31. Beko had its own site cited for 1 of 21, and Sharp and Tornado for none of 26 between them.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">In the answer text, B.TECH was named in 13 answers, Carrefour in 12 and Jumia in 10. Independent repair businesses were cited in 6 answers, mostly about gas cookers. Review sites written for other countries, such as Which? and Consumer Reports, appeared in 3.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">AI Mode reads brand sites and social media. Brand domains appeared on its answer pages in 37 of 60 answers. A Facebook post or group appeared among its source cards in 39, and a YouTube video in 10. Across all 280 ChatGPT answers in this study, neither Facebook nor YouTube was cited once. The two engines are measured differently: ChatGPT&#8217;s sources come from its links and chips, AI Mode&#8217;s from the source cards and domains on its page.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">AI Mode also read the maker of two of its leading brands. Elaraby Group, which makes Sharp and Tornado in Egypt, appeared as a source in 25 of AI Mode&#8217;s 60 answers, against 1 or 2 of 60 in each Arabic ChatGPT run. Tornado and Sharp were also the brands AI Mode recommended far more often than ChatGPT. We did not test whether one causes the other.<\/p>\r\n\r\n\r\n\r\n<figure class=\"fig-wrap\" data-fig=\"fig-sources.svg\"><svg class=\"fig\" xmlns=\"http:\/\/www.w3.org\/2000\/svg\" viewBox=\"0 0 900 428\" role=\"img\" aria-label=\"Shop sites are cited about as often as brand sites. ChatGPT, first run: answers (of 60) that cite at least one source of each type, as a link or a source chip.\">\r\n  <style>\r\n    .fig { --ink:#12203a; --mut:#5a6b86; --grid:#dfe5ee; --bg:#ffffff; --glob:#1f5f8b; --home:#b4532a; --pale:#d6dde8; }\r\n    @media (prefers-color-scheme: dark) {\r\n      .fig { --ink:#e8edf5; --mut:#9dabc2; --grid:#2b3648; --bg:#0f1520; --glob:#6fb2dd; --home:#e0855c; --pale:#38445a; }\r\n    }\r\n    .fig text { font-family: ui-sans-serif, system-ui, \"Segoe UI\", Arial, sans-serif; fill: var(--ink); }\r\n    .fig .mut { fill: var(--mut); }\r\n    .fig .grid { stroke: var(--grid); stroke-width: 1; }\r\n    .fig .ttl { font-weight: 700; }\r\n  <\/style>\r\n<rect width=\"900\" height=\"428\" fill=\"var(--bg)\"\/>\r\n<text x=\"24\" y=\"32\" class=\"ttl\" font-size=\"19\">Shop sites are cited about as often as brand sites<\/text>\r\n<text x=\"24\" y=\"54\" class=\"mut\" font-size=\"13\">ChatGPT, first run: answers (of 60) that cite at least one source of each type, as a link or a source chip.<\/text>\r\n\r\n<text x=\"318\" y=\"114\" font-size=\"13\" text-anchor=\"end\">Shop or dealer<\/text>\r\n<rect x=\"330\" y=\"100\" width=\"242.8\" height=\"22\" fill=\"var(--glob)\"\/>\r\n<text x=\"580.8\" y=\"115\" font-size=\"12\">31<\/text>\r\n<text x=\"318\" y=\"148\" font-size=\"13\" text-anchor=\"end\">Brand&#8217;s own site<\/text>\r\n<rect x=\"330\" y=\"134\" width=\"211.5\" height=\"22\" fill=\"var(--glob)\"\/>\r\n<text x=\"549.5\" y=\"149\" font-size=\"12\">27<\/text>\r\n<text x=\"318\" y=\"182\" font-size=\"13\" text-anchor=\"end\">Egyptian buying guide<\/text>\r\n<rect x=\"330\" y=\"168\" width=\"70.5\" height=\"22\" fill=\"var(--glob)\"\/>\r\n<text x=\"408.5\" y=\"183\" font-size=\"12\">9<\/text>\r\n<text x=\"318\" y=\"216\" font-size=\"13\" text-anchor=\"end\">Independent repair business<\/text>\r\n<rect x=\"330\" y=\"202\" width=\"47.0\" height=\"22\" fill=\"var(--home)\"\/>\r\n<text x=\"385.0\" y=\"217\" font-size=\"12\">6<\/text>\r\n<text x=\"318\" y=\"250\" font-size=\"13\" text-anchor=\"end\">News, research, government<\/text>\r\n<rect x=\"330\" y=\"236\" width=\"47.0\" height=\"22\" fill=\"var(--glob)\"\/>\r\n<text x=\"385.0\" y=\"251\" font-size=\"12\">6<\/text>\r\n<text x=\"318\" y=\"284\" font-size=\"13\" text-anchor=\"end\">Price comparison<\/text>\r\n<rect x=\"330\" y=\"270\" width=\"39.2\" height=\"22\" fill=\"var(--glob)\"\/>\r\n<text x=\"377.2\" y=\"285\" font-size=\"12\">5<\/text>\r\n<text x=\"318\" y=\"318\" font-size=\"13\" text-anchor=\"end\">Brand-named site run by a third party<\/text>\r\n<rect x=\"330\" y=\"304\" width=\"31.3\" height=\"22\" fill=\"var(--home)\"\/>\r\n<text x=\"369.3\" y=\"319\" font-size=\"12\">4<\/text>\r\n<text x=\"318\" y=\"352\" font-size=\"13\" text-anchor=\"end\">Review site for another country<\/text>\r\n<rect x=\"330\" y=\"338\" width=\"23.5\" height=\"22\" fill=\"var(--glob)\"\/>\r\n<text x=\"361.5\" y=\"353\" font-size=\"12\">3<\/text>\r\n<text x=\"24\" y=\"414\" font-size=\"11\" class=\"mut\">Source: VOCTOS, 60 logged-out ChatGPT answers captured in Egypt, 13 to 14 October 2026.<\/text>\r\n<\/svg><figcaption>Shop sites are cited about as often as brand sites. ChatGPT, first run: answers (of 60) that cite at least one source of each type, as a link or a source chip.<\/figcaption><\/figure>\r\n\r\n\r\n\r\n<h2 class=\"wp-block-heading\">Do short search-style prompts change the answer?<\/h2>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">They change who comes first, and on ChatGPT they came with ads. We typed 40 prompts the way Egyptians type a search box, four per category, on both engines.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">The plain prompt behaves like the full question. \u00ab\u0627\u0641\u0636\u0644 \u062b\u0644\u0627\u062c\u0629 \u0641\u064a \u0645\u0635\u0631\u00bb and \u00ab\u0625\u064a\u0647 \u0623\u062d\u0633\u0646 \u0645\u0627\u0631\u0643\u0629 \u062b\u0644\u0627\u062c\u0629 \u0641\u064a \u0645\u0635\u0631\u061f\u00bb shared the same first pick in 7 of 10 categories on ChatGPT, and 8 of 10 on AI Mode.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">The attribute does not. Adding inverter, no-frost, a size or \u00ab\u0627\u0642\u062a\u0635\u0627\u062f\u064a\u00bb changed the first pick in 20 of 30 comparisons on ChatGPT and 17 of 30 on AI Mode. The economy version is where Egyptian brands lead: an Egyptian brand came first in 4 of 10 economy prompts on ChatGPT and 7 of 10 on AI Mode, against none of the plain prompts on either.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">The two engines agreed on the first pick for only 12 of the 40 identical prompts.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\"><strong>The ads.<\/strong> 24 of 40 search-style answers on ChatGPT carried a sponsored product unit, with prices from Egyptian sellers, served through Criteo&#8217;s ad network. Full questions rarely did: 1 of 60 in the first run, 0 in the repeat, 1 in English and 3 in formal Arabic, the run just before. We did not register this before the search-style run, and the prompts ran as a separate block, so prompt form and time of day are tangled. Treat it as an observation. It still matters: a buyer who types like a search box can see an advertiser&#8217;s product under the answer.<\/p>\r\n\r\n\r\n\r\n<h2 class=\"wp-block-heading\">What drives the variation?<\/h2>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">Mostly the fit between brand and question, and least of all the engine or language. Across all five runs of the 60 questions and 59 brands recommended at least once, the split looks like this.<\/p>\r\n\r\n\r\n\r\n<figure class=\"wp-block-table table-scroll\"><table><thead><tr><th>Source of variation<\/th><th>Share<\/th><\/tr><\/thead><tbody><tr><td>Which brand fits which question (brand by question)<\/td><td>42.0%<\/td><\/tr><tr><td>Answer-to-answer noise<\/td><td>39.2%<\/td><\/tr><tr><td>The brand itself, across all questions<\/td><td>17.0%<\/td><\/tr><tr><td>The run, language or engine, brand by brand<\/td><td>1.2%<\/td><\/tr><\/tbody><\/table><\/figure>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">The biggest share is fit: a dishwasher question brings Beko and Bosch, a freezer question brings Fresh. The question on its own, apart from which brands fit it, explains almost nothing. The second share is noise, the part no design variable explains; with one answer per cell, it also absorbs any effect of run that differs by brand and question. The run, language or engine shifts a brand&#8217;s overall rate by little. What changing language or engine does is make single answers less alike. Answer-level noise is 32.9% of the spread between two Arabic runs, 34.0% between Egyptian and formal Arabic, 43.6% between ChatGPT and AI Mode and 46.1% between Arabic and English.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">A 2026 study of European brands found the query language explained more of the variation than the brand (<a href=\"https:\/\/arxiv.org\/abs\/2607.13304\" target=\"_blank\" rel=\"noopener\">arXiv<\/a>). That study measured sentiment, not recommendations, so the numbers are not comparable. In our data, for recommendations, the fit between brand and question is the largest share.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\"><strong>For a brand, this means the unit of work is the question, on each engine. Find the questions where you fit and are missing.<\/strong><\/p>\r\n\r\n\r\n\r\n<figure class=\"fig-wrap\" data-fig=\"fig-variance.svg\"><svg class=\"fig\" xmlns=\"http:\/\/www.w3.org\/2000\/svg\" viewBox=\"0 0 900 400\" role=\"img\" aria-label=\"Fit with the question matters more than the engine or the language. Share of the variation in whether a brand is recommended, by source. Other small components are left out.\">\r\n  <style>\r\n    .fig { --ink:#12203a; --mut:#5a6b86; --grid:#dfe5ee; --bg:#ffffff; --glob:#1f5f8b; --home:#b4532a; --pale:#d6dde8; }\r\n    @media (prefers-color-scheme: dark) {\r\n      .fig { --ink:#e8edf5; --mut:#9dabc2; --grid:#2b3648; --bg:#0f1520; --glob:#6fb2dd; --home:#e0855c; --pale:#38445a; }\r\n    }\r\n    .fig text { font-family: ui-sans-serif, system-ui, \"Segoe UI\", Arial, sans-serif; fill: var(--ink); }\r\n    .fig .mut { fill: var(--mut); }\r\n    .fig .grid { stroke: var(--grid); stroke-width: 1; }\r\n    .fig .ttl { font-weight: 700; }\r\n  <\/style>\r\n<rect width=\"900\" height=\"400\" fill=\"var(--bg)\"\/>\r\n<text x=\"24\" y=\"32\" class=\"ttl\" font-size=\"19\">Fit with the question matters more than the engine or the language<\/text>\r\n<text x=\"24\" y=\"54\" class=\"mut\" font-size=\"13\">Share of the variation in whether a brand is recommended, by source. Other small components are left out.<\/text>\r\n\r\n<rect x=\"24\" y=\"70\" width=\"12\" height=\"12\" fill=\"var(--home)\" opacity=\"1.0\"\/><text x=\"42\" y=\"80\" font-size=\"12\">brand fits the question<\/text>\r\n<rect x=\"284\" y=\"70\" width=\"12\" height=\"12\" fill=\"var(--glob)\" opacity=\"1.0\"\/><text x=\"302\" y=\"80\" font-size=\"12\">the brand itself<\/text>\r\n<rect x=\"24\" y=\"90\" width=\"12\" height=\"12\" fill=\"var(--ink)\" opacity=\"1.0\"\/><text x=\"42\" y=\"100\" font-size=\"12\">run, language or engine<\/text>\r\n<rect x=\"284\" y=\"90\" width=\"12\" height=\"12\" fill=\"var(--pale)\" opacity=\"1.0\"\/><text x=\"302\" y=\"100\" font-size=\"12\">answer-to-answer noise<\/text>\r\n<text x=\"218\" y=\"140\" font-size=\"13\" text-anchor=\"end\">Arabic, asked twice<\/text>\r\n<rect x=\"230.0\" y=\"124\" width=\"330.1\" height=\"30\" fill=\"var(--home)\" opacity=\"1.0\"\/>\r\n<text x=\"395.1\" y=\"144\" font-size=\"11\" text-anchor=\"middle\" style=\"fill:var(--bg)\">52%<\/text>\r\n<rect x=\"560.1\" y=\"124\" width=\"85.7\" height=\"30\" fill=\"var(--glob)\" opacity=\"1.0\"\/>\r\n<text x=\"603.0\" y=\"144\" font-size=\"11\" text-anchor=\"middle\" style=\"fill:var(--bg)\">14%<\/text>\r\n<rect x=\"645.8\" y=\"124\" width=\"1.9\" height=\"30\" fill=\"var(--ink)\" opacity=\"1.0\"\/>\r\n<rect x=\"647.7\" y=\"124\" width=\"207.3\" height=\"30\" fill=\"var(--pale)\" opacity=\"1.0\"\/>\r\n<text x=\"751.3\" y=\"144\" font-size=\"11\" text-anchor=\"middle\" style=\"fill:var(--ink)\">33%<\/text>\r\n<text x=\"218\" y=\"184\" font-size=\"13\" text-anchor=\"end\">Egyptian vs formal Arabic<\/text>\r\n<rect x=\"230.0\" y=\"168\" width=\"316.9\" height=\"30\" fill=\"var(--home)\" opacity=\"1.0\"\/>\r\n<text x=\"388.4\" y=\"188\" font-size=\"11\" text-anchor=\"middle\" style=\"fill:var(--bg)\">50%<\/text>\r\n<rect x=\"546.9\" y=\"168\" width=\"93.2\" height=\"30\" fill=\"var(--glob)\" opacity=\"1.0\"\/>\r\n<text x=\"593.5\" y=\"188\" font-size=\"11\" text-anchor=\"middle\" style=\"fill:var(--bg)\">15%<\/text>\r\n<rect x=\"640.1\" y=\"168\" width=\"1.3\" height=\"30\" fill=\"var(--ink)\" opacity=\"1.0\"\/>\r\n<rect x=\"641.4\" y=\"168\" width=\"214.2\" height=\"30\" fill=\"var(--pale)\" opacity=\"1.0\"\/>\r\n<text x=\"748.5\" y=\"188\" font-size=\"11\" text-anchor=\"middle\" style=\"fill:var(--ink)\">34%<\/text>\r\n<text x=\"218\" y=\"228\" font-size=\"13\" text-anchor=\"end\">ChatGPT vs AI Mode<\/text>\r\n<rect x=\"230.0\" y=\"212\" width=\"237.5\" height=\"30\" fill=\"var(--home)\" opacity=\"1.0\"\/>\r\n<text x=\"348.8\" y=\"232\" font-size=\"11\" text-anchor=\"middle\" style=\"fill:var(--bg)\">38%<\/text>\r\n<rect x=\"467.5\" y=\"212\" width=\"97.7\" height=\"30\" fill=\"var(--glob)\" opacity=\"1.0\"\/>\r\n<text x=\"516.3\" y=\"232\" font-size=\"11\" text-anchor=\"middle\" style=\"fill:var(--bg)\">16%<\/text>\r\n<rect x=\"565.2\" y=\"212\" width=\"8.8\" height=\"30\" fill=\"var(--ink)\" opacity=\"1.0\"\/>\r\n<rect x=\"574.0\" y=\"212\" width=\"274.7\" height=\"30\" fill=\"var(--pale)\" opacity=\"1.0\"\/>\r\n<text x=\"711.3\" y=\"232\" font-size=\"11\" text-anchor=\"middle\" style=\"fill:var(--ink)\">44%<\/text>\r\n<text x=\"218\" y=\"272\" font-size=\"13\" text-anchor=\"end\">Arabic vs English<\/text>\r\n<rect x=\"230.0\" y=\"256\" width=\"253.9\" height=\"30\" fill=\"var(--home)\" opacity=\"1.0\"\/>\r\n<text x=\"356.9\" y=\"276\" font-size=\"11\" text-anchor=\"middle\" style=\"fill:var(--bg)\">40%<\/text>\r\n<rect x=\"483.9\" y=\"256\" width=\"78.1\" height=\"30\" fill=\"var(--glob)\" opacity=\"1.0\"\/>\r\n<text x=\"523.0\" y=\"276\" font-size=\"11\" text-anchor=\"middle\" style=\"fill:var(--bg)\">12%<\/text>\r\n<rect x=\"562.0\" y=\"256\" width=\"5.7\" height=\"30\" fill=\"var(--ink)\" opacity=\"1.0\"\/>\r\n<rect x=\"567.7\" y=\"256\" width=\"290.4\" height=\"30\" fill=\"var(--pale)\" opacity=\"1.0\"\/>\r\n<text x=\"712.9\" y=\"276\" font-size=\"11\" text-anchor=\"middle\" style=\"fill:var(--ink)\">46%<\/text>\r\n<text x=\"218\" y=\"316\" font-size=\"13\" text-anchor=\"end\">All five runs<\/text>\r\n<rect x=\"230.0\" y=\"300\" width=\"264.6\" height=\"30\" fill=\"var(--home)\" opacity=\"1.0\"\/>\r\n<text x=\"362.3\" y=\"320\" font-size=\"11\" text-anchor=\"middle\" style=\"fill:var(--bg)\">42%<\/text>\r\n<rect x=\"494.6\" y=\"300\" width=\"107.1\" height=\"30\" fill=\"var(--glob)\" opacity=\"1.0\"\/>\r\n<text x=\"548.1\" y=\"320\" font-size=\"11\" text-anchor=\"middle\" style=\"fill:var(--bg)\">17%<\/text>\r\n<rect x=\"601.7\" y=\"300\" width=\"7.6\" height=\"30\" fill=\"var(--ink)\" opacity=\"1.0\"\/>\r\n<rect x=\"609.3\" y=\"300\" width=\"247.0\" height=\"30\" fill=\"var(--pale)\" opacity=\"1.0\"\/>\r\n<text x=\"732.7\" y=\"320\" font-size=\"11\" text-anchor=\"middle\" style=\"fill:var(--ink)\">39%<\/text>\r\n<text x=\"24\" y=\"386\" font-size=\"11\" class=\"mut\">Source: VOCTOS, the same 60 questions asked five ways, 300 answers, 13 to 14 October 2026. Generalizability study, brand x question x run.<\/text>\r\n<\/svg><figcaption>Fit with the question matters more than the engine or the language. Share of the variation in whether a brand is recommended, by source. Other small components are left out.<\/figcaption><\/figure>\r\n\r\n\r\n\r\n<h2 class=\"wp-block-heading\">Why do some brands never appear?<\/h2>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">The clearest pattern is demand. Brands that Egyptians search for, by appliance and name, are the brands both engines recommend. Across the 25 panel brands the rankings correlate at 0.91 in the first ChatGPT run and 0.93 on AI Mode. The five brands ChatGPT never named in Egyptian Arabic rank 20th, 21st, 22nd, 24th and 25th of 25 on search demand.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">The scale is stark. Searches for Fresh with an appliance word run to about 44,000 a month in Semrush&#8217;s Egypt data. Electrolux gets about 50, Whirlpool 470, Gree 910, Black+Decker 1,410 and Kenwood 1,590. Fresh draws 880 times Electrolux&#8217;s volume. Inside each category the match is looser: the three most searched brands and the three most recommended share 1.9 names on average.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">Some brands sit off the line. For anyone planning work on a brand they are the most useful part of our data, because they show where search interest exists and recommendations do not follow.<\/p>\r\n\r\n\r\n\r\n<ul class=\"wp-block-list\"><li><strong>Philips is over-recommended.<\/strong> It draws about 680 category searches a month in our measure, yet was recommended in 4 vacuum answers. Our measure counts only &#8220;appliance plus brand&#8221; searches, so people searching for a Philips model number are not in it.<\/li><li><strong>Haier is under-recommended.<\/strong> It draws 9,880 searches, more than Kiriazi&#8217;s 9,790, but was recommended 3 times against Kiriazi&#8217;s 7. Its Egyptian site is in English only, and Semrush credits it with 268 ranking keywords in Egypt.<\/li><li><strong>Panasonic shows the same gap.<\/strong> It draws 9,390 searches, 8,310 of them for Panasonic vacuum cleaners, and was recommended 3 times. Vacuum cleaners were the category where search and recommendations agreed least, at 0.32.<\/li><\/ul>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">The usual technical explanations do not separate the visible from the missing. No robots.txt file we could read blocks OpenAI&#8217;s crawlers, including those of the three missing brands that have a site; Beko&#8217;s and Panasonic&#8217;s sites refused our script altogether. All five missing brands have English and Arabic Wikipedia articles, while Fresh has only an Arabic one. Fresh&#8217;s English sitemap lists 186 URLs, none of them under \/products\/.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">Four of the missing five also lack an Egyptian site. Electrolux runs a regional site, and in Egypt it is present through brands it owns: Zanussi, recommended 9 times, and Olympic Electric, recommended 4 times. We found no Egyptian site for Whirlpool or Gree, and Black+Decker&#8217;s nearest is in the UAE. Kenwood has an English page for Egypt; its Arabic address returns an error. Neither Kenwood, a kitchen-appliance brand, nor Black+Decker, a tool brand with a small-appliance line, makes most of the large appliances in our set, and only 2 of our 10 categories are small appliances.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">Low demand is not a rule, though. Philips ranks 23rd of 25 on our demand measure and was still recommended 4 times. <strong>What we can say is narrower: every brand that neither engine recommended in Egyptian Arabic was one Egyptians rarely search for by appliance and name.<\/strong><\/p>\r\n\r\n\r\n\r\n<figure class=\"fig-wrap\" data-fig=\"fig-demand.svg\"><svg class=\"fig\" xmlns=\"http:\/\/www.w3.org\/2000\/svg\" viewBox=\"0 0 900 643\" role=\"img\" aria-label=\"ChatGPT names the brands Egyptians already search for. 25 brands sorted by search demand. Left: share of Egyptian searches for &#x27;&lt;appliance&gt; &lt;brand&gt;&#x27;. Right: share of recommendations. Spearman 0.914.\">\r\n  <style>\r\n    .fig { --ink:#12203a; --mut:#5a6b86; --grid:#dfe5ee; --bg:#ffffff; --glob:#1f5f8b; --home:#b4532a; --pale:#d6dde8; }\r\n    @media (prefers-color-scheme: dark) {\r\n      .fig { --ink:#e8edf5; --mut:#9dabc2; --grid:#2b3648; --bg:#0f1520; --glob:#6fb2dd; --home:#e0855c; --pale:#38445a; }\r\n    }\r\n    .fig text { font-family: ui-sans-serif, system-ui, \"Segoe UI\", Arial, sans-serif; fill: var(--ink); }\r\n    .fig .mut { fill: var(--mut); }\r\n    .fig .grid { stroke: var(--grid); stroke-width: 1; }\r\n    .fig .ttl { font-weight: 700; }\r\n  <\/style>\r\n<rect width=\"900\" height=\"643\" fill=\"var(--bg)\"\/>\r\n<text x=\"24\" y=\"32\" class=\"ttl\" font-size=\"19\">ChatGPT names the brands Egyptians already search for<\/text>\r\n<text x=\"24\" y=\"54\" class=\"mut\" font-size=\"13\">25 brands sorted by search demand. Left: share of Egyptian searches for &#x27;&lt;appliance&gt; &lt;brand&gt;&#x27;. Right: share of recommendations. Spearman 0.914.<\/text>\r\n\r\n<text x=\"380\" y=\"92\" font-size=\"12\" class=\"mut\" text-anchor=\"end\">share of searches<\/text><text x=\"520\" y=\"92\" font-size=\"12\" class=\"mut\">share of ChatGPT recommendations<\/text><circle cx=\"640\" cy=\"72\" r=\"6\" fill=\"var(--glob)\"\/><text x=\"652\" y=\"76\" font-size=\"12\">global brand<\/text><circle cx=\"750\" cy=\"72\" r=\"6\" fill=\"var(--home)\"\/><text x=\"762\" y=\"76\" font-size=\"12\">made in Egypt<\/text>\r\n<rect x=\"86.5\" y=\"115\" width=\"303.5\" height=\"13\" fill=\"var(--home)\" opacity=\"0.45\"\/>\r\n<rect x=\"510.0\" y=\"115\" width=\"338.3\" height=\"13\" fill=\"var(--home)\"\/>\r\n<text x=\"450\" y=\"126\" font-size=\"12\" text-anchor=\"middle\">Fresh<\/text>\r\n<text x=\"80.5\" y=\"126\" font-size=\"10\" class=\"mut\" text-anchor=\"end\">14.7%<\/text>\r\n<text x=\"854.3\" y=\"126\" font-size=\"10\" class=\"mut\">16.4%<\/text>\r\n<rect x=\"152.9\" y=\"134\" width=\"237.1\" height=\"13\" fill=\"var(--glob)\" opacity=\"0.45\"\/>\r\n<rect x=\"510.0\" y=\"134\" width=\"196.4\" height=\"13\" fill=\"var(--glob)\"\/>\r\n<text x=\"450\" y=\"145\" font-size=\"12\" text-anchor=\"middle\">LG<\/text>\r\n<text x=\"146.9\" y=\"145\" font-size=\"10\" class=\"mut\" text-anchor=\"end\">11.5%<\/text>\r\n<text x=\"712.4\" y=\"145\" font-size=\"10\" class=\"mut\">9.5%<\/text>\r\n<rect x=\"173.6\" y=\"153\" width=\"216.4\" height=\"13\" fill=\"var(--glob)\" opacity=\"0.45\"\/>\r\n<rect x=\"510.0\" y=\"153\" width=\"152.8\" height=\"13\" fill=\"var(--glob)\"\/>\r\n<text x=\"450\" y=\"164\" font-size=\"12\" text-anchor=\"middle\">Sharp<\/text>\r\n<text x=\"167.6\" y=\"164\" font-size=\"10\" class=\"mut\" text-anchor=\"end\">10.5%<\/text>\r\n<text x=\"668.8\" y=\"164\" font-size=\"10\" class=\"mut\">7.4%<\/text>\r\n<rect x=\"191.0\" y=\"172\" width=\"199.0\" height=\"13\" fill=\"var(--glob)\" opacity=\"0.45\"\/>\r\n<rect x=\"510.0\" y=\"172\" width=\"229.2\" height=\"13\" fill=\"var(--glob)\"\/>\r\n<text x=\"450\" y=\"183\" font-size=\"12\" text-anchor=\"middle\">Toshiba<\/text>\r\n<text x=\"185.0\" y=\"183\" font-size=\"10\" class=\"mut\" text-anchor=\"end\">9.6%<\/text>\r\n<text x=\"745.2\" y=\"183\" font-size=\"10\" class=\"mut\">11.1%<\/text>\r\n<rect x=\"226.0\" y=\"191\" width=\"164.0\" height=\"13\" fill=\"var(--glob)\" opacity=\"0.45\"\/>\r\n<rect x=\"510.0\" y=\"191\" width=\"229.2\" height=\"13\" fill=\"var(--glob)\"\/>\r\n<text x=\"450\" y=\"202\" font-size=\"12\" text-anchor=\"middle\">Beko<\/text>\r\n<text x=\"220.0\" y=\"202\" font-size=\"10\" class=\"mut\" text-anchor=\"end\">8.0%<\/text>\r\n<text x=\"745.2\" y=\"202\" font-size=\"10\" class=\"mut\">11.1%<\/text>\r\n<rect x=\"275.7\" y=\"210\" width=\"114.3\" height=\"13\" fill=\"var(--home)\" opacity=\"0.45\"\/>\r\n<rect x=\"510.0\" y=\"210\" width=\"131.0\" height=\"13\" fill=\"var(--home)\"\/>\r\n<text x=\"450\" y=\"221\" font-size=\"12\" text-anchor=\"middle\">Tornado<\/text>\r\n<text x=\"269.7\" y=\"221\" font-size=\"10\" class=\"mut\" text-anchor=\"end\">5.5%<\/text>\r\n<text x=\"647.0\" y=\"221\" font-size=\"10\" class=\"mut\">6.3%<\/text>\r\n<rect x=\"287.4\" y=\"229\" width=\"102.6\" height=\"13\" fill=\"var(--glob)\" opacity=\"0.45\"\/>\r\n<rect x=\"510.0\" y=\"229\" width=\"120.0\" height=\"13\" fill=\"var(--glob)\"\/>\r\n<text x=\"450\" y=\"240\" font-size=\"12\" text-anchor=\"middle\">Samsung<\/text>\r\n<text x=\"281.4\" y=\"240\" font-size=\"10\" class=\"mut\" text-anchor=\"end\">5.0%<\/text>\r\n<text x=\"636.0\" y=\"240\" font-size=\"10\" class=\"mut\">5.8%<\/text>\r\n<rect x=\"290.5\" y=\"248\" width=\"99.5\" height=\"13\" fill=\"var(--glob)\" opacity=\"0.45\"\/>\r\n<rect x=\"510.0\" y=\"248\" width=\"120.0\" height=\"13\" fill=\"var(--glob)\"\/>\r\n<text x=\"450\" y=\"259\" font-size=\"12\" text-anchor=\"middle\">Bosch<\/text>\r\n<text x=\"284.5\" y=\"259\" font-size=\"10\" class=\"mut\" text-anchor=\"end\">4.8%<\/text>\r\n<text x=\"636.0\" y=\"259\" font-size=\"10\" class=\"mut\">5.8%<\/text>\r\n<rect x=\"305.1\" y=\"267\" width=\"84.9\" height=\"13\" fill=\"var(--glob)\" opacity=\"0.45\"\/>\r\n<rect x=\"510.0\" y=\"267\" width=\"76.4\" height=\"13\" fill=\"var(--glob)\"\/>\r\n<text x=\"450\" y=\"278\" font-size=\"12\" text-anchor=\"middle\">Midea<\/text>\r\n<text x=\"299.1\" y=\"278\" font-size=\"10\" class=\"mut\" text-anchor=\"end\">4.1%<\/text>\r\n<text x=\"592.4\" y=\"278\" font-size=\"10\" class=\"mut\">3.7%<\/text>\r\n<rect x=\"321.9\" y=\"286\" width=\"68.1\" height=\"13\" fill=\"var(--glob)\" opacity=\"0.45\"\/>\r\n<rect x=\"510.0\" y=\"286\" width=\"32.7\" height=\"13\" fill=\"var(--glob)\"\/>\r\n<text x=\"450\" y=\"297\" font-size=\"12\" text-anchor=\"middle\">Haier<\/text>\r\n<text x=\"315.9\" y=\"297\" font-size=\"10\" class=\"mut\" text-anchor=\"end\">3.3%<\/text>\r\n<text x=\"548.7\" y=\"297\" font-size=\"10\" class=\"mut\">1.6%<\/text>\r\n<rect x=\"322.5\" y=\"305\" width=\"67.5\" height=\"13\" fill=\"var(--home)\" opacity=\"0.45\"\/>\r\n<rect x=\"510.0\" y=\"305\" width=\"76.4\" height=\"13\" fill=\"var(--home)\"\/>\r\n<text x=\"450\" y=\"316\" font-size=\"12\" text-anchor=\"middle\">Kiriazi<\/text>\r\n<text x=\"316.5\" y=\"316\" font-size=\"10\" class=\"mut\" text-anchor=\"end\">3.3%<\/text>\r\n<text x=\"592.4\" y=\"316\" font-size=\"10\" class=\"mut\">3.7%<\/text>\r\n<rect x=\"325.2\" y=\"324\" width=\"64.8\" height=\"13\" fill=\"var(--glob)\" opacity=\"0.45\"\/>\r\n<rect x=\"510.0\" y=\"324\" width=\"32.7\" height=\"13\" fill=\"var(--glob)\"\/>\r\n<text x=\"450\" y=\"335\" font-size=\"12\" text-anchor=\"middle\">Panasonic<\/text>\r\n<text x=\"319.2\" y=\"335\" font-size=\"10\" class=\"mut\" text-anchor=\"end\">3.1%<\/text>\r\n<text x=\"548.7\" y=\"335\" font-size=\"10\" class=\"mut\">1.6%<\/text>\r\n<rect x=\"327.5\" y=\"343\" width=\"62.5\" height=\"13\" fill=\"var(--glob)\" opacity=\"0.45\"\/>\r\n<rect x=\"510.0\" y=\"343\" width=\"98.2\" height=\"13\" fill=\"var(--glob)\"\/>\r\n<text x=\"450\" y=\"354\" font-size=\"12\" text-anchor=\"middle\">Zanussi<\/text>\r\n<text x=\"321.5\" y=\"354\" font-size=\"10\" class=\"mut\" text-anchor=\"end\">3.0%<\/text>\r\n<text x=\"614.2\" y=\"354\" font-size=\"10\" class=\"mut\">4.8%<\/text>\r\n<rect x=\"334.0\" y=\"362\" width=\"56.0\" height=\"13\" fill=\"var(--glob)\" opacity=\"0.45\"\/>\r\n<rect x=\"510.0\" y=\"362\" width=\"54.6\" height=\"13\" fill=\"var(--glob)\"\/>\r\n<text x=\"450\" y=\"373\" font-size=\"12\" text-anchor=\"middle\">Ariston<\/text>\r\n<text x=\"328.0\" y=\"373\" font-size=\"10\" class=\"mut\" text-anchor=\"end\">2.7%<\/text>\r\n<text x=\"570.6\" y=\"373\" font-size=\"10\" class=\"mut\">2.6%<\/text>\r\n<rect x=\"339.2\" y=\"381\" width=\"50.8\" height=\"13\" fill=\"var(--home)\" opacity=\"0.45\"\/>\r\n<rect x=\"510.0\" y=\"381\" width=\"43.7\" height=\"13\" fill=\"var(--home)\"\/>\r\n<text x=\"450\" y=\"392\" font-size=\"12\" text-anchor=\"middle\">Universal<\/text>\r\n<text x=\"333.2\" y=\"392\" font-size=\"10\" class=\"mut\" text-anchor=\"end\">2.5%<\/text>\r\n<text x=\"559.7\" y=\"392\" font-size=\"10\" class=\"mut\">2.1%<\/text>\r\n<rect x=\"350.1\" y=\"400\" width=\"39.9\" height=\"13\" fill=\"var(--glob)\" opacity=\"0.45\"\/>\r\n<rect x=\"510.0\" y=\"400\" width=\"32.7\" height=\"13\" fill=\"var(--glob)\"\/>\r\n<text x=\"450\" y=\"411\" font-size=\"12\" text-anchor=\"middle\">Carrier<\/text>\r\n<text x=\"344.1\" y=\"411\" font-size=\"10\" class=\"mut\" text-anchor=\"end\">1.9%<\/text>\r\n<text x=\"548.7\" y=\"411\" font-size=\"10\" class=\"mut\">1.6%<\/text>\r\n<rect x=\"354.9\" y=\"419\" width=\"35.1\" height=\"13\" fill=\"var(--home)\" opacity=\"0.45\"\/>\r\n<rect x=\"510.0\" y=\"419\" width=\"10.9\" height=\"13\" fill=\"var(--home)\"\/>\r\n<text x=\"450\" y=\"430\" font-size=\"12\" text-anchor=\"middle\">White Whale<\/text>\r\n<text x=\"348.9\" y=\"430\" font-size=\"10\" class=\"mut\" text-anchor=\"end\">1.7%<\/text>\r\n<text x=\"526.9\" y=\"430\" font-size=\"10\" class=\"mut\">0.5%<\/text>\r\n<rect x=\"356.9\" y=\"438\" width=\"33.1\" height=\"13\" fill=\"var(--home)\" opacity=\"0.45\"\/>\r\n<rect x=\"510.0\" y=\"438\" width=\"32.7\" height=\"13\" fill=\"var(--home)\"\/>\r\n<text x=\"450\" y=\"449\" font-size=\"12\" text-anchor=\"middle\">Unionaire<\/text>\r\n<text x=\"350.9\" y=\"449\" font-size=\"10\" class=\"mut\" text-anchor=\"end\">1.6%<\/text>\r\n<text x=\"548.7\" y=\"449\" font-size=\"10\" class=\"mut\">1.6%<\/text>\r\n<rect x=\"362.1\" y=\"457\" width=\"27.9\" height=\"13\" fill=\"var(--glob)\" opacity=\"0.45\"\/>\r\n<rect x=\"510.0\" y=\"457\" width=\"10.9\" height=\"13\" fill=\"var(--glob)\"\/>\r\n<text x=\"450\" y=\"468\" font-size=\"12\" text-anchor=\"middle\">Hoover<\/text>\r\n<text x=\"356.1\" y=\"468\" font-size=\"10\" class=\"mut\" text-anchor=\"end\">1.4%<\/text>\r\n<text x=\"526.9\" y=\"468\" font-size=\"10\" class=\"mut\">0.5%<\/text>\r\n<rect x=\"379.0\" y=\"476\" width=\"11.0\" height=\"13\" fill=\"var(--glob)\" opacity=\"0.45\"\/>\r\n<rect x=\"510.0\" y=\"476\" width=\"0.0\" height=\"13\" fill=\"var(--glob)\"\/>\r\n<text x=\"450\" y=\"487\" font-size=\"12\" text-anchor=\"middle\">Kenwood<\/text>\r\n<text x=\"373.0\" y=\"487\" font-size=\"10\" class=\"mut\" text-anchor=\"end\">0.5%<\/text>\r\n<text x=\"516.0\" y=\"487\" font-size=\"10\" class=\"mut\">never named<\/text>\r\n<rect x=\"380.3\" y=\"495\" width=\"9.7\" height=\"13\" fill=\"var(--glob)\" opacity=\"0.45\"\/>\r\n<rect x=\"510.0\" y=\"495\" width=\"0.0\" height=\"13\" fill=\"var(--glob)\"\/>\r\n<text x=\"450\" y=\"506\" font-size=\"12\" text-anchor=\"middle\">Black+Decker<\/text>\r\n<text x=\"374.3\" y=\"506\" font-size=\"10\" class=\"mut\" text-anchor=\"end\">0.5%<\/text>\r\n<text x=\"516.0\" y=\"506\" font-size=\"10\" class=\"mut\">never named<\/text>\r\n<rect x=\"383.7\" y=\"514\" width=\"6.3\" height=\"13\" fill=\"var(--glob)\" opacity=\"0.45\"\/>\r\n<rect x=\"510.0\" y=\"514\" width=\"0.0\" height=\"13\" fill=\"var(--glob)\"\/>\r\n<text x=\"450\" y=\"525\" font-size=\"12\" text-anchor=\"middle\">Gree<\/text>\r\n<text x=\"377.7\" y=\"525\" font-size=\"10\" class=\"mut\" text-anchor=\"end\">0.3%<\/text>\r\n<text x=\"516.0\" y=\"525\" font-size=\"10\" class=\"mut\">never named<\/text>\r\n<rect x=\"385.3\" y=\"533\" width=\"4.7\" height=\"13\" fill=\"var(--glob)\" opacity=\"0.45\"\/>\r\n<rect x=\"510.0\" y=\"533\" width=\"43.7\" height=\"13\" fill=\"var(--glob)\"\/>\r\n<text x=\"450\" y=\"544\" font-size=\"12\" text-anchor=\"middle\">Philips<\/text>\r\n<text x=\"379.3\" y=\"544\" font-size=\"10\" class=\"mut\" text-anchor=\"end\">0.2%<\/text>\r\n<text x=\"559.7\" y=\"544\" font-size=\"10\" class=\"mut\">2.1%<\/text>\r\n<rect x=\"386.8\" y=\"552\" width=\"3.2\" height=\"13\" fill=\"var(--glob)\" opacity=\"0.45\"\/>\r\n<rect x=\"510.0\" y=\"552\" width=\"0.0\" height=\"13\" fill=\"var(--glob)\"\/>\r\n<text x=\"450\" y=\"563\" font-size=\"12\" text-anchor=\"middle\">Whirlpool<\/text>\r\n<text x=\"380.8\" y=\"563\" font-size=\"10\" class=\"mut\" text-anchor=\"end\">0.2%<\/text>\r\n<text x=\"516.0\" y=\"563\" font-size=\"10\" class=\"mut\">never named<\/text>\r\n<rect x=\"389.7\" y=\"571\" width=\"0.3\" height=\"13\" fill=\"var(--glob)\" opacity=\"0.45\"\/>\r\n<rect x=\"510.0\" y=\"571\" width=\"0.0\" height=\"13\" fill=\"var(--glob)\"\/>\r\n<text x=\"450\" y=\"582\" font-size=\"12\" text-anchor=\"middle\">Electrolux<\/text>\r\n<text x=\"383.7\" y=\"582\" font-size=\"10\" class=\"mut\" text-anchor=\"end\">0.0%<\/text>\r\n<text x=\"516.0\" y=\"582\" font-size=\"10\" class=\"mut\">never named<\/text>\r\n<text x=\"24\" y=\"629\" font-size=\"11\" class=\"mut\">Source: VOCTOS, 60 logged-out ChatGPT answers captured in Egypt, 13 to 14 October 2026. Searches: Semrush Egypt database, Arabic and Latin spellings, ten categories.<\/text>\r\n<\/svg><figcaption>ChatGPT names the brands Egyptians already search for. 25 brands sorted by search demand. Left: share of Egyptian searches for &#x27;&lt;appliance&gt; &lt;brand&gt;&#x27;. Right: share of recommendations. Spearman 0.914.<\/figcaption><\/figure>\r\n\r\n\r\n\r\n<h2 class=\"wp-block-heading\">Do Egyptian brands lose to imports?<\/h2>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">Not overall. Made-in-Egypt brands took 62 of 208 recommendation slots in the first run (29.8%), and 40 of the 52 answers that recommended brands included at least one of them. Fresh alone was the first pick 14 times.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">A split appears when you look at what the user asked, though we added this breakdown after seeing the data. In the first run, ChatGPT never put an Egyptian brand first for &#8220;best brand&#8221;, in 10 answers, although Egyptian brands still held 12 of the 53 places in those lists. For &#8220;best value&#8221; it put one first in 5 of 10, and for &#8220;Egyptian or imported&#8221; and &#8220;where to buy&#8221; in 5 of 6 each. The repeat run softened the &#8220;best brand&#8221; result to 2 of 10, and English matched it at 2 of 10.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">We did not code how each brand was described, so we cannot say Egyptian brands were praised for price. What the order shows is narrower. When the question is &#8220;which is best&#8221;, the first name a buyer reads is usually a global brand: 18 times in the 20 Arabic answers across both runs. That count includes the TV answers read as computer monitors, and it treats Sharp, made in Egypt by Elaraby, as global. The search-style prompts show the same split from the other side: &#8220;economy&#8221; puts Egyptian brands first, &#8220;best&#8221; does not.<\/p>\r\n\r\n\r\n\r\n<figure class=\"fig-wrap\" data-fig=\"fig-qtype.svg\"><svg class=\"fig\" xmlns=\"http:\/\/www.w3.org\/2000\/svg\" viewBox=\"0 0 900 400\" role=\"img\" aria-label=\"First run: Egyptian brands came first for value, never for &#x27;best&#x27;. Answers whose first pick is a made-in-Egypt brand, by question type, first run (breakdown added after the data). n = answers that recommended a brand.\">\r\n  <style>\r\n    .fig { --ink:#12203a; --mut:#5a6b86; --grid:#dfe5ee; --bg:#ffffff; --glob:#1f5f8b; --home:#b4532a; --pale:#d6dde8; }\r\n    @media (prefers-color-scheme: dark) {\r\n      .fig { --ink:#e8edf5; --mut:#9dabc2; --grid:#2b3648; --bg:#0f1520; --glob:#6fb2dd; --home:#e0855c; --pale:#38445a; }\r\n    }\r\n    .fig text { font-family: ui-sans-serif, system-ui, \"Segoe UI\", Arial, sans-serif; fill: var(--ink); }\r\n    .fig .mut { fill: var(--mut); }\r\n    .fig .grid { stroke: var(--grid); stroke-width: 1; }\r\n    .fig .ttl { font-weight: 700; }\r\n  <\/style>\r\n<rect width=\"900\" height=\"400\" fill=\"var(--bg)\"\/>\r\n<text x=\"24\" y=\"32\" class=\"ttl\" font-size=\"19\">First run: Egyptian brands came first for value, never for &#x27;best&#x27;<\/text>\r\n<text x=\"24\" y=\"54\" class=\"mut\" font-size=\"13\">Answers whose first pick is a made-in-Egypt brand, by question type, first run (breakdown added after the data). n = answers that recommended a brand.<\/text>\r\n\r\n<text x=\"248\" y=\"115\" font-size=\"13\" text-anchor=\"end\">Best brand<\/text>\r\n<rect x=\"260\" y=\"100\" width=\"560\" height=\"26\" fill=\"var(--pale)\"\/>\r\n<text x=\"268.0\" y=\"116\" font-size=\"12\">0 of 10<\/text>\r\n<text x=\"248\" y=\"155\" font-size=\"13\" text-anchor=\"end\">Lasts longest<\/text>\r\n<rect x=\"260\" y=\"140\" width=\"560\" height=\"26\" fill=\"var(--pale)\"\/>\r\n<rect x=\"260\" y=\"140\" width=\"56.0\" height=\"26\" fill=\"var(--home)\"\/>\r\n<text x=\"324.0\" y=\"156\" font-size=\"12\">1 of 10<\/text>\r\n<text x=\"248\" y=\"195\" font-size=\"13\" text-anchor=\"end\">Spare parts and service<\/text>\r\n<rect x=\"260\" y=\"180\" width=\"560\" height=\"26\" fill=\"var(--pale)\"\/>\r\n<rect x=\"260\" y=\"180\" width=\"112.0\" height=\"26\" fill=\"var(--home)\"\/>\r\n<text x=\"380.0\" y=\"196\" font-size=\"12\">2 of 10<\/text>\r\n<text x=\"248\" y=\"235\" font-size=\"13\" text-anchor=\"end\">Best value<\/text>\r\n<rect x=\"260\" y=\"220\" width=\"560\" height=\"26\" fill=\"var(--pale)\"\/>\r\n<rect x=\"260\" y=\"220\" width=\"280.0\" height=\"26\" fill=\"var(--home)\"\/>\r\n<text x=\"548.0\" y=\"236\" font-size=\"12\">5 of 10<\/text>\r\n<text x=\"248\" y=\"275\" font-size=\"13\" text-anchor=\"end\">Egyptian or imported<\/text>\r\n<rect x=\"260\" y=\"260\" width=\"560\" height=\"26\" fill=\"var(--pale)\"\/>\r\n<rect x=\"260\" y=\"260\" width=\"466.7\" height=\"26\" fill=\"var(--home)\"\/>\r\n<text x=\"734.7\" y=\"276\" font-size=\"12\">5 of 6<\/text>\r\n<text x=\"248\" y=\"315\" font-size=\"13\" text-anchor=\"end\">Where to buy<\/text>\r\n<rect x=\"260\" y=\"300\" width=\"560\" height=\"26\" fill=\"var(--pale)\"\/>\r\n<rect x=\"260\" y=\"300\" width=\"466.7\" height=\"26\" fill=\"var(--home)\"\/>\r\n<text x=\"734.7\" y=\"316\" font-size=\"12\">5 of 6<\/text>\r\n<text x=\"24\" y=\"386\" font-size=\"11\" class=\"mut\">Source: VOCTOS, 60 logged-out ChatGPT answers captured in Egypt, 13 to 14 October 2026.<\/text>\r\n<\/svg><figcaption>First run: Egyptian brands came first for value, never for &#x27;best&#x27;. Answers whose first pick is a made-in-Egypt brand, by question type, first run (breakdown added after the data). n = answers that recommended a brand.<\/figcaption><\/figure>\r\n\r\n\r\n\r\n<h2 class=\"wp-block-heading\">Which sites answer the spare-parts and service questions?<\/h2>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">Ten questions asked which brand has service and spare parts available in Egypt. Nine first-run answers cited sources. Five cited only brands&#8217; own sites, and three cited sites that use a brand&#8217;s name but are run by someone else.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">Toshiba shows both sides. Its Egyptian appliance site now carries a Midea Electrics Egypt copyright; Elaraby announced in December 2025 that its Toshiba agreements would end with that year. The official site gives a hotline of 19111. In two answers ChatGPT cited that site and quoted 19111 correctly. In two others the Toshiba service claim came from toshibaeg.com, toshiba-egy.com and toshibae.com. One of those says it is the only officially approved Toshiba service in Egypt and advertises 19028; another says on the page that it is not the official site. In one of these answers ChatGPT itself warned that some independent repair sites call themselves &#8220;approved&#8221;.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">The air-conditioner answer went further. Under &#8220;website&#8221;, it linked Sharp to sharp-egypt.com and Carrier to carrier-egy.com. The first calls itself Sharp&#8217;s official site, but Sharp&#8217;s Egyptian maker, Elaraby, runs elarabygroup.com, whose homepage does not link to it, and we found nothing tying sharp-egypt.com to Elaraby. The second describes itself in its footer as an independent authorised distributor. Carrier&#8217;s Egyptian joint venture, Miraco, runs miraco.com.eg, which no answer cited.<\/p>\r\n\r\n\r\n\r\n<figure class=\"wp-block-table table-scroll\"><table><thead><tr><th>Domain cited<\/th><th>Brand in the name<\/th><th>What the site says about itself<\/th><\/tr><\/thead><tbody><tr><td>toshiba-egy.com<\/td><td>Toshiba<\/td><td>&#8220;the only officially approved Toshiba service in Egypt&#8221;, hotline 19028<\/td><\/tr><tr><td>toshibaeg.com<\/td><td>Toshiba<\/td><td>calls its number &#8220;the approved Toshiba maintenance number&#8221;; footer names a separate owner<\/td><\/tr><tr><td>toshibae.com<\/td><td>Toshiba<\/td><td>&#8220;we are not the official site or the agent&#8221;<\/td><\/tr><tr><td>sharp-egypt.com<\/td><td>Sharp<\/td><td>calls itself Sharp&#8217;s official site and agent<\/td><\/tr><tr><td>carrier-egy.com<\/td><td>Carrier<\/td><td>&#8220;independent authorized distributor&#8221;<\/td><\/tr><tr><td>carrier-cairo.com<\/td><td>Carrier<\/td><td>&#8220;Carrier&#8217;s official site&#8221; in its title<\/td><\/tr><tr><td>olympic-electric-eg.com<\/td><td>Olympic Electric<\/td><td>&#8220;an independent centre&#8221;<\/td><\/tr><\/tbody><\/table><\/figure>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">toshiba-egy.com uses the same claim, with the brand name swapped, and the same 19028 number as whirlpool-egypt.com, a repair site we found while auditing Whirlpool. The last row of the table comes from a &#8220;lasts longest&#8221; answer, where Olympic Electric&#8217;s recommendation was backed by two independent repair sites rather than the brand&#8217;s own store.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">We are not saying any of these businesses misled anyone. The point is narrower. The numbers ChatGPT wrote into its answers were the brands&#8217; own, but a user who clicks one of these sources reaches that business&#8217;s number, not the brand&#8217;s.<\/p>\r\n\r\n\r\n\r\n<h2 class=\"wp-block-heading\">What do the product pictures in ChatGPT&#8217;s answers show?<\/h2>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">This was not pre-registered. Forty first-run answers showed a strip of product pictures, 134 images in all. We read each image&#8217;s title for a market signal: Egypt, another country, or none.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">53 images pointed to Egypt, through Egyptian shops, prices in pounds or Egyptian model codes. 26 pointed to another country (19.4%), and 13 of those had titles in a language other than Arabic or English. 19 of the 40 strips (47.5%) contained at least one foreign listing. They included a Bosch vacuum priced in Danish kroner, a Samsung TV from a Mexican Sam&#8217;s Club and a Sharp microwave priced for Bangladesh.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">The answer text quoted prices in Egyptian pounds. In 19 of 40 strips, at least one picture beside it pointed to another country. A reader who clicks one can land in a store that does not ship to Egypt.<\/p>\r\n\r\n\r\n\r\n<h2 class=\"wp-block-heading\">What else did we see?<\/h2>\r\n\r\n\r\n\r\n<ul class=\"wp-block-list\"><li><strong>Every answer asked for more detail.<\/strong> All 60 first-run answers closed by asking about budget, size or type, most through a tap-to-answer form and a few in plain text.<\/li><li><strong>ChatGPT assumed a location.<\/strong> Eight first-run answers referred to Qalyubia governorate, by recommending shops in Qalyub or Banha, offering its towns as location choices, or mentioning service there. No question named a city or governorate.<\/li><li><strong>Eight of the ten &#8220;lasts longest&#8221; answers said no brand can be guaranteed<\/strong> not to break.<\/li><li><strong>One sponsored unit appeared in 60 first-run answers.<\/strong> It sat under an air-conditioner question and advertised an installer with a US phone number.<\/li><\/ul>\r\n\r\n\r\n\r\n<h2 class=\"wp-block-heading\">The brand audit, brand by brand<\/h2>\r\n\r\n\r\n\r\n<figure class=\"wp-block-table table-scroll\"><table><thead><tr><th>Brand<\/th><th>Recommended in<\/th><th>Searches a month<\/th><th>Egyptian site<\/th><th>Arabic<\/th><th>Product markup<\/th><th>Wikipedia EN \/ AR<\/th><th>Answers citing own site<\/th><\/tr><\/thead><tbody><tr><td>Fresh<\/td><td>31<\/td><td>44,000<\/td><td>fresh.com.eg<\/td><td>yes<\/td><td>yes on be.fresh.com.eg; no product URLs in sitemap<\/td><td>no \/ yes<\/td><td>6<\/td><\/tr><tr><td>Toshiba<\/td><td>21<\/td><td>28,850<\/td><td>toshiba-lifestyle.com\/eg<\/td><td>yes<\/td><td>no<\/td><td>yes \/ yes<\/td><td>4<\/td><\/tr><tr><td>Beko<\/td><td>21<\/td><td>23,780<\/td><td>beko.com\/eg-ar<\/td><td>yes<\/td><td>not reached<\/td><td>yes \/ yes<\/td><td>1<\/td><\/tr><tr><td>LG<\/td><td>18<\/td><td>34,380<\/td><td>lg.com\/eg_ar<\/td><td>yes<\/td><td>not reached<\/td><td>yes \/ yes<\/td><td>8<\/td><\/tr><tr><td>Sharp<\/td><td>14<\/td><td>31,380<\/td><td>elarabygroup.com<\/td><td>yes<\/td><td>yes<\/td><td>yes \/ yes<\/td><td>1<\/td><\/tr><tr><td>Tornado<\/td><td>12<\/td><td>16,570<\/td><td>elarabygroup.com<\/td><td>yes<\/td><td>yes<\/td><td>no \/ no<\/td><td>1<\/td><\/tr><tr><td>Samsung<\/td><td>11<\/td><td>14,880<\/td><td>samsung.com\/eg<\/td><td>yes<\/td><td>yes<\/td><td>yes \/ yes<\/td><td>4<\/td><\/tr><tr><td>Bosch<\/td><td>11<\/td><td>14,420<\/td><td>bosch-home.com\/eg<\/td><td>yes<\/td><td>yes<\/td><td>yes \/ not found<\/td><td>6<\/td><\/tr><tr><td>Zanussi<\/td><td>9<\/td><td>9,060<\/td><td>zanussi.com.eg<\/td><td>yes<\/td><td>not reached<\/td><td>yes \/ yes<\/td><td>5<\/td><\/tr><tr><td>Midea<\/td><td>7<\/td><td>12,310<\/td><td>midea.com\/eg<\/td><td>yes<\/td><td>yes<\/td><td>yes \/ yes<\/td><td>1<\/td><\/tr><tr><td>Kiriazi<\/td><td>7<\/td><td>9,790<\/td><td>kiriazi.com<\/td><td>yes<\/td><td>yes<\/td><td>no \/ yes<\/td><td>2<\/td><\/tr><tr><td>Ariston<\/td><td>5<\/td><td>8,120<\/td><td>ariston.com\/ar-eg<\/td><td>yes<\/td><td>not reached<\/td><td>yes \/ yes<\/td><td>2<\/td><\/tr><tr><td>Universal<\/td><td>4<\/td><td>7,370<\/td><td>universalgroup.org<\/td><td>English only<\/td><td>not reached<\/td><td>no \/ no<\/td><td>0<\/td><\/tr><tr><td>Philips<\/td><td>4<\/td><td>680<\/td><td>philips.com.eg<\/td><td>yes<\/td><td>not reached<\/td><td>yes \/ yes<\/td><td>1<\/td><\/tr><tr><td>Carrier<\/td><td>3<\/td><td>5,790<\/td><td>miraco.com.eg<\/td><td>yes<\/td><td>not reached<\/td><td>yes \/ yes<\/td><td>0<\/td><\/tr><tr><td>Panasonic<\/td><td>3<\/td><td>9,390<\/td><td>panasonic.com\/eg<\/td><td>blocked our check<\/td><td>blocked our check<\/td><td>yes \/ yes<\/td><td>0<\/td><\/tr><tr><td>Haier<\/td><td>3<\/td><td>9,880<\/td><td>haier.com\/eg<\/td><td>English only<\/td><td>no<\/td><td>yes \/ yes<\/td><td>1<\/td><\/tr><tr><td>Unionaire<\/td><td>3<\/td><td>4,800<\/td><td>unionaire.com<\/td><td>yes<\/td><td>not reached<\/td><td>no \/ yes<\/td><td>1<\/td><\/tr><tr><td>Hoover<\/td><td>1<\/td><td>4,050<\/td><td>none found<\/td><td><\/td><td><\/td><td>yes \/ not checked<\/td><td>0<\/td><\/tr><tr><td>White Whale<\/td><td>1<\/td><td>5,090<\/td><td>whitewhale.com.eg<\/td><td>yes<\/td><td>not reached<\/td><td>no \/ no<\/td><td>0<\/td><\/tr><tr><td>Electrolux<\/td><td>0<\/td><td>50<\/td><td>regional site only<\/td><td><\/td><td><\/td><td>yes \/ yes<\/td><td>0<\/td><\/tr><tr><td>Whirlpool<\/td><td>0<\/td><td>470<\/td><td>none found<\/td><td><\/td><td><\/td><td>yes \/ yes<\/td><td>0<\/td><\/tr><tr><td>Gree<\/td><td>0<\/td><td>910<\/td><td>none found<\/td><td><\/td><td><\/td><td>yes \/ yes<\/td><td>0<\/td><\/tr><tr><td>Black+Decker<\/td><td>0<\/td><td>1,410<\/td><td>UAE site only<\/td><td><\/td><td><\/td><td>yes \/ yes<\/td><td>0<\/td><\/tr><tr><td>Kenwood<\/td><td>0<\/td><td>1,590<\/td><td>kenwoodworld.com\/en-eg<\/td><td>English only<\/td><td>not reached<\/td><td>yes \/ yes<\/td><td>0<\/td><\/tr><\/tbody><\/table><\/figure>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">&#8220;Recommended in&#8221; counts the first ChatGPT run. &#8220;Not reached&#8221; means our script found no product page to test; it does not mean the markup is missing. Semrush traffic to each brand&#8217;s Egyptian site also tracks recommendations, more loosely, at 0.65 across the 19 sites with data.<\/p>\r\n\r\n\r\n\r\n<h2 class=\"wp-block-heading\">What we got wrong, and what surprised us<\/h2>\r\n\r\n\r\n\r\n<ul class=\"wp-block-list\"><li><strong>A first-run claim did not survive the repeat.<\/strong> After wave 1 we wrote that ChatGPT never put an Egyptian brand first for &#8220;best brand&#8221;. Asked again, it did in 2 of 10. One run is not enough to say &#8220;never&#8221;.<\/li><li><strong>We expected formal Arabic to behave like English.<\/strong> On slots it did not: register changed the lists about as much as asking twice. Egyptian brands did come first less often in it, so the answer is not a clean no.<\/li><li><strong>We did not expect Facebook.<\/strong> AI Mode cited brand sites too, but Facebook posts or groups appeared in 39 of 60 of its answers, a source ChatGPT never used.<\/li><li><strong>We expected search-style prompts to be a cosmetic change.<\/strong> On ChatGPT they came with sponsored product ads in 24 of 40 answers.<\/li><li><strong>We expected the engine to shift brands&#8217; overall rates more than it did.<\/strong> Brand by brand, the run, language or engine explained about 1% of the variation, against 42% for the fit between brand and question. What engine and language did change was single answers, which became less alike.<\/li><\/ul>\r\n\r\n\r\n\r\n<h2 class=\"wp-block-heading\">What this does not prove<\/h2>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\"><strong>Demand and recommendations move together; we did not show one causes the other.<\/strong> Brands that sell more are searched more and stocked more, and we did not separate those three. The demand comparison was also added after the data.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\"><strong>A few runs per question.<\/strong> Each question was asked three times in Egyptian Arabic or formal Arabic on ChatGPT, once in English and once on AI Mode. That is enough to see how much lists move, not to estimate any single brand&#8217;s rate precisely, which is why we report counts with intervals rather than a league table. Sources, service sites and product pictures come from the first run only.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\"><strong>Logged out, one machine, one day.<\/strong> Logged-in users, paid plans and the mobile app may see different answers, and engines change without notice. The waves ran within 12 hours of each other.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\"><strong>AI Mode sources are read from the page.<\/strong> Its links are Google redirects, so we counted the domains and site names AI Mode displayed, not the pages behind them.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\"><strong>Search demand is an estimate.<\/strong> Semrush volumes are modelled, and brands searched by name alone, like Philips, are undercounted by an &#8220;appliance plus brand&#8221; measure.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\"><strong>The audit is partial.<\/strong> We reached a product page on 8 of the 20 sites we tested, plus one Fresh page on a separate subdomain, and two sites refused scripted requests. Universal&#8217;s site comes from a job listing, not the company&#8217;s own channels.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\"><strong>Several coders, checked.<\/strong> A second coder, working without sight of our coding, recoded 12 randomly drawn first-run answers holding 38 recommendations and 8 mentions, and matched every one, in the same order. Later waves were coded by other blind coders under the same rules, so differences between runs include any difference between coders. Agreement between those coders was not measured.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\"><strong>Conflict of interest.<\/strong> VOCTOS sells the work this study points to. None of the 25 panel brands is a VOCTOS client at the time of publication.<\/p>\r\n\r\n\r\n\r\n<h2 class=\"wp-block-heading\">What to do with it<\/h2>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">The useful question is not whether ChatGPT is allowed to read your website. No robots.txt we could read stops it. The question is whether the pages each engine does read say what you want said.<\/p>\r\n\r\n\r\n\r\n<ol class=\"wp-block-list\"><li><strong>Fix what shops say about you.<\/strong> Shop sites as a group, led by B.TECH, Jumia and Carrefour Egypt, were cited about as often as brands&#8217; own sites. Full Arabic specifications, current prices and correct model names on those sites cost little and sit where ChatGPT looks.<\/li><li><strong>Treat Facebook as a source, not just a channel.<\/strong> AI Mode cited Facebook groups and posts in 39 of 60 answers. What owners say about your brand in those groups is now part of the answer.<\/li><li><strong>Own the service answer.<\/strong> Publish one Arabic service page on your own domain with the hotline in text, the service centres and the warranty terms. Toshiba has one and still had third-party sites cited in two of the four service answers that named it.<\/li><li><strong>Win the attribute, not just the category.<\/strong> Inverter, no-frost, size and economy prompts changed the first pick in most comparisons. Make sure your pages and listings say which of those you are.<\/li><li><strong>Cover English too, if you are an Egyptian brand.<\/strong> Your share fell by a quarter when the same questions were asked in English.<\/li><li><strong>Measure over many answers, not one.<\/strong> A single screenshot proves little; the first pick changed in about half of repeated questions. Track Recommendation Rate across repeated runs, languages and both engines.<\/li><li><strong>Then check the basics.<\/strong> An Egyptian site in Arabic, product pages in your sitemap and Product markup did not separate winners from losers here, but they keep your own pages usable when an engine does look.<\/li><\/ol>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">If your brand has real demand and few recommendations, like Haier or Panasonic in this data, the gap may be in the sources, which you can work on. If demand itself is low, as for the five missing brands, start there.<\/p>\r\n\r\n\r\n\r\n<h2 class=\"wp-block-heading\">How this compares with earlier studies<\/h2>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">The closest study in subject is 5W Research&#8217;s 2026 index of major kitchen appliances in the US, which asked 46 prompts three times on four assistants (<a href=\"https:\/\/www.5wpr.com\/ai-visibility-index\/major-appliances-ai-visibility-index-2026\/\" target=\"_blank\" rel=\"noopener\">5W<\/a>). It also found brands&#8217; own sites rarely cited, and that the leading brands overlap too much to rank precisely. Where it found review publishers like Consumer Reports cited most, Egyptian answers leaned on Egyptian shops, one Egyptian guide and, on AI Mode, Facebook. 5W&#8217;s general methodology page describes a different process from the one in the appliance report, so we compare only with figures the report itself states.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">Ahrefs studied 75,000 brands in Google&#8217;s AI Overviews and found brand web mentions correlated with AI visibility at 0.664, and branded search volume at 0.392 (<a href=\"https:\/\/ahrefs.com\/blog\/ai-overview-brand-correlation\/\" target=\"_blank\" rel=\"noopener\">Ahrefs<\/a>). Our 0.91 is much tighter. It comes from one market, one category and 25 brands, so the samples are not comparable. The direction agrees: being talked about and searched for correlates more strongly with being named than site mechanics do.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">On language, a study of Gulf hotel recommendations found English and Arabic shortlists largely alike but their sources less so (<a href=\"https:\/\/martawarren.com\/gcc-ai-discovery-gap\/\" target=\"_blank\" rel=\"noopener\">Marta Warren and Rankscale<\/a>). Our English answers moved against local brands, as the multilingual preprint above predicts. Research on language models has found that they favour local brands once the user&#8217;s country is stated, except for electronics (<a href=\"https:\/\/aclanthology.org\/2024.emnlp-main.707\" target=\"_blank\" rel=\"noopener\">ACL Anthology<\/a>). Our questions all named Egypt, so global brands coming first for &#8220;best&#8221; fits that finding only if appliances behave like electronics.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">On service numbers, security researchers have documented AI answers that repeat scam support numbers planted on third-party pages (<a href=\"https:\/\/aurascape.ai\/resources\/auralabs-research\/llm-search-poisoning-fake-support-numbers\/\" target=\"_blank\" rel=\"noopener\">Aurascape<\/a>). We found no fake numbers in ChatGPT&#8217;s own words, but the same exposure in its sources: brand-named sites that are not the brand.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">Our own study of <a href=\"https:\/\/www.voctos.com\/research\/chatgpt-local-recommendations-egypt\/\">local businesses in Egypt<\/a> found that most of ChatGPT&#8217;s local picks were shown with map data. Product brands have no map card, and here shops, guides, service sites and, on AI Mode, Facebook filled that role instead.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">As far as we could find, no earlier study measured AI brand recommendations in Egyptian Arabic, compared it with formal Arabic, or covered the Egyptian market.<\/p>\r\n\r\n\r\n\r\n<h2 class=\"wp-block-heading\">Corrections and right of reply<\/h2>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">If you represent a brand named here and think we recorded what an engine said about it wrongly, <a href=\"https:\/\/www.voctos.com\/contact\/\">contact us<\/a>. We will check your point against the captured answer, which is public in the data files below, and correct the page and the data if we got it wrong, with a dated note of the change. We will not remove a brand that an engine did recommend, because the study records what buyers were told.<\/p>\r\n\r\n\r\n\r\n<h2 class=\"wp-block-heading\">Questions people ask about this<\/h2>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\"><strong>Which appliance brand does ChatGPT recommend most in Egypt?<\/strong><\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">Fresh. In 60 Egyptian Arabic appliance questions asked on 13 October 2026, ChatGPT recommended Fresh in 31 answers and in 9 of the 10 categories. Asked again it recommended Fresh in 28, in English in 27 and in Modern Standard Arabic in 35. Toshiba, LG and Beko follow it.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\"><strong>Does Google AI Mode recommend the same appliance brands as ChatGPT in Egypt?<\/strong><\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">The same leaders in a different order. AI Mode recommended Tornado in 31 of 60 answers and Fresh in 29, and named 49 brands in all. It put an Egyptian brand first less often than ChatGPT, in 9 of 49 answers, and cited Facebook posts or groups in 39 of 60.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\"><strong>Why doesn&#8217;t ChatGPT recommend my appliance brand in Egypt?<\/strong><\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">In our data the strongest pattern was search demand. The five brands ChatGPT never named in 120 answers to the full Egyptian Arabic questions, Electrolux, Whirlpool, Gree, Black+Decker and Kenwood, are among the six least searched in Egypt, and four have no Egyptian website. Robots.txt rules and Wikipedia did not separate visible brands from missing ones.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\"><strong>Does ChatGPT prefer Egyptian or imported appliance brands?<\/strong><\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">Neither overall, in Arabic. Egyptian brands took 29.8% of recommendations in the first run and 33.0% in the repeat, but 24.7% when the same questions were asked in English. They were the first pick for &#8220;best brand&#8221; in only 2 of 20 Arabic answers, but came first in 4 of 10 &#8220;economy&#8221; prompts on ChatGPT and 7 of 10 on AI Mode.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\"><strong>Does ChatGPT show ads when you ask about appliances in Egypt?<\/strong><\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">Rarely for full questions, often for short ones. Sponsored product units appeared under 24 of 40 search-style prompts such as \u00ab\u0627\u0641\u0636\u0644 \u062a\u0643\u064a\u064a\u0641 \u0627\u0646\u0641\u0631\u062a\u0631 \u0641\u064a \u0645\u0635\u0631\u00bb, against 3 of 60 full questions in Modern Standard Arabic asked just before them.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\"><strong>Does ChatGPT recommend the same brands every time?<\/strong><\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">Not exactly. Asked the same Arabic question twice, it kept the same first pick in 24 of 47 pairs and the identical brand list in 11 of 54. But the leading brands kept appearing: Fresh was recommended in 31 and then 28 of 60 answers, LG in 18 and 16.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\"><strong>Can I trust the service hotlines ChatGPT gives for appliance brands in Egypt?<\/strong><\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">In our data, the hotlines ChatGPT wrote into its answers were the brands&#8217; own, such as Toshiba Egypt&#8217;s 19111. The sources behind them were another matter: in 3 of 10 service answers ChatGPT cited third-party sites that use a brand&#8217;s name, one of them advertising a different Toshiba number. Check any number against the brand&#8217;s own site before you call.<\/p>\r\n\r\n\r\n\r\n<h2 class=\"wp-block-heading\">Data and citation<\/h2>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">Three files hold everything behind this page: the <a href=\"https:\/\/www.voctos.com\/wp-content\/uploads\/2026\/10\/voctos-egypt-appliances-chatgpt-coded-2026-10-v2.csv\">coded answers<\/a> from all seven runs, with every brand recommended, mentioned or cautioned and every source domain; the <a href=\"https:\/\/www.voctos.com\/wp-content\/uploads\/2026\/10\/voctos-egypt-appliances-brand-audit-2026-10.csv\">brand audit table<\/a>; and the <a href=\"https:\/\/www.voctos.com\/wp-content\/uploads\/2026\/10\/voctos-egypt-appliances-chatgpt-answers-2026-10-v2.csv\">full text of all 380 answers<\/a> with every link, source chip and product picture title. Anyone can recode the answers from the third file.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">Cite as: VOCTOS (2026). <em>Which appliance brands ChatGPT and Google AI Mode recommend in Egypt (380 answers studied).<\/em> Captured 13 to 14 October 2026. https:\/\/www.voctos.com\/research\/chatgpt-appliance-brands-egypt\/<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">Revised on 14 October 2026. Added the formal Arabic, Google AI Mode and search-style runs, intervals that resample whole questions, the variance split, cautions and a corrections policy. No first-run code changed.<\/p>\r\n\r\n\r\n\r\n<p class=\"wp-block-paragraph\">If you want to see what ChatGPT and AI Mode say about your own brand, our <a href=\"https:\/\/www.voctos.com\/ai-search-audit\/\">AI search audit<\/a> runs the questions your buyers ask. The work that follows sits with <a href=\"https:\/\/www.voctos.com\/geo\/chatgpt-optimization\/\">ChatGPT optimization<\/a> and <a href=\"https:\/\/www.voctos.com\/geo\/gemini-ai-mode\/\">Gemini and AI Mode optimization<\/a> for the pages they read, and <a href=\"https:\/\/www.voctos.com\/seo\/ecommerce\/\">e-commerce SEO<\/a> for your product pages.<\/p>\r\n\r\n\r\n\r\n\r\n<\/div><\/div>\r\n\r\n\n\r\n<script type=\"application\/ld+json\">\r\n{\r\n  \"@context\": \"https:\/\/schema.org\",\r\n  \"@graph\": [\r\n    {\r\n      \"@type\": \"ScholarlyArticle\",\r\n      \"@id\": \"https:\/\/www.voctos.com\/research\/chatgpt-appliance-brands-egypt\/#article\",\r\n      \"url\": \"https:\/\/www.voctos.com\/research\/chatgpt-appliance-brands-egypt\/\",\r\n      \"name\": \"Which appliance brands ChatGPT and Google AI Mode recommend in Egypt (380 answers studied)\",\r\n      \"headline\": \"Which appliance brands ChatGPT and Google AI Mode recommend in Egypt (380 answers studied)\",\r\n      \"description\": \"Original measurement of the home appliance brands logged-out ChatGPT and Google AI Mode recommend in Egypt: 60 Egyptian Arabic questions across 10 categories asked five ways (twice in Egyptian Arabic, once in English and once in Modern Standard Arabic on ChatGPT, once on AI Mode) plus 40 search-style prompts on each engine, 380 answers in all, with a same-day audit of 25 brands. Fresh was recommended in 31 of 60 first-run answers; recommendations track Egyptian search demand (Spearman 0.914 on ChatGPT, 0.928 on AI Mode); made-in-Egypt brands took 33.0% of slots in Arabic and 24.7% in English; AI Mode cited Facebook in 39 of 60 answers; 24 of 40 search-style ChatGPT prompts carried sponsored product units.\",\r\n      \"inLanguage\": \"en\",\r\n      \"datePublished\": \"2026-10-14\",\r\n      \"dateModified\": \"2026-10-14\",\r\n      \"isAccessibleForFree\": true,\r\n      \"creativeWorkStatus\": \"Published\",\r\n      \"author\": {\r\n        \"@id\": \"https:\/\/www.voctos.com\/#organization\"\r\n      },\r\n      \"contributor\": {\r\n        \"@type\": \"Person\",\r\n        \"@id\": \"https:\/\/www.voctos.com\/team\/beshoy-adel\/#person\",\r\n        \"name\": \"Beshoy Adel\",\r\n        \"jobTitle\": \"Head of SEO and GEO\",\r\n        \"url\": \"https:\/\/www.voctos.com\/team\/beshoy-adel\/\",\r\n        \"worksFor\": {\r\n          \"@id\": \"https:\/\/www.voctos.com\/#organization\"\r\n        }\r\n      },\r\n      \"publisher\": {\r\n        \"@id\": \"https:\/\/www.voctos.com\/#organization\"\r\n      },\r\n      \"about\": [\r\n        {\r\n          \"@type\": \"Thing\",\r\n          \"name\": \"ChatGPT\"\r\n        },\r\n        {\r\n          \"@type\": \"Thing\",\r\n          \"name\": \"Home appliances\"\r\n        },\r\n        {\r\n          \"@type\": \"Thing\",\r\n          \"name\": \"Generative engine optimization\"\r\n        },\r\n        {\r\n          \"@type\": \"Thing\",\r\n          \"name\": \"Brand visibility in AI search\"\r\n        },\r\n        {\r\n          \"@type\": \"Place\",\r\n          \"name\": \"Egypt\"\r\n        }\r\n      ],\r\n      \"mentions\": [\r\n        {\r\n          \"@type\": \"Brand\",\r\n          \"name\": \"Fresh\"\r\n        },\r\n        {\r\n          \"@type\": \"Brand\",\r\n          \"name\": \"Toshiba\"\r\n        },\r\n        {\r\n          \"@type\": \"Brand\",\r\n          \"name\": \"Beko\"\r\n        },\r\n        {\r\n          \"@type\": \"Brand\",\r\n          \"name\": \"LG\"\r\n        },\r\n        {\r\n          \"@type\": \"Brand\",\r\n          \"name\": \"Sharp\"\r\n        },\r\n        {\r\n          \"@type\": \"Brand\",\r\n          \"name\": \"Tornado\"\r\n        },\r\n        {\r\n          \"@type\": \"Brand\",\r\n          \"name\": \"Unionaire\"\r\n        },\r\n        {\r\n          \"@type\": \"Brand\",\r\n          \"name\": \"Electrolux\"\r\n        },\r\n        {\r\n          \"@type\": \"Brand\",\r\n          \"name\": \"Whirlpool\"\r\n        }\r\n      ],\r\n      \"keywords\": \"ChatGPT brand recommendations, home appliances Egypt, AI search visibility, generative engine optimization, Fresh, Toshiba, Beko, B.TECH\",\r\n      \"citation\": [\r\n        {\r\n          \"@type\": \"Article\",\r\n          \"name\": \"An Analysis of AI Overview Brand Visibility Factors (75K Brands Studied)\",\r\n          \"url\": \"https:\/\/ahrefs.com\/blog\/ai-overview-brand-correlation\/\",\r\n          \"publisher\": {\r\n            \"@type\": \"Organization\",\r\n            \"name\": \"Ahrefs\"\r\n          }\r\n        },\r\n        {\r\n          \"@type\": \"Article\",\r\n          \"name\": \"Major Kitchen Appliances AI Visibility Index 2026\",\r\n          \"url\": \"https:\/\/www.5wpr.com\/ai-visibility-index\/major-appliances-ai-visibility-index-2026\/\",\r\n          \"publisher\": {\r\n            \"@type\": \"Organization\",\r\n            \"name\": \"5W Research\"\r\n          }\r\n        },\r\n        {\r\n          \"@type\": \"ScholarlyArticle\",\r\n          \"name\": \"The Language of the Question Selects the Market\",\r\n          \"url\": \"https:\/\/arxiv.org\/abs\/2608.30052\"\r\n        },\r\n        {\r\n          \"@type\": \"Article\",\r\n          \"name\": \"AI recommendation lists rarely repeat\",\r\n          \"url\": \"https:\/\/searchengineland.com\/ai-recommendation-lists-rarely-repeat-study-468076\",\r\n          \"publisher\": {\r\n            \"@type\": \"Organization\",\r\n            \"name\": \"Search Engine Land\"\r\n          }\r\n        },\r\n        {\r\n          \"@type\": \"ScholarlyArticle\",\r\n          \"name\": \"Where Does the Noise Come From?\",\r\n          \"url\": \"https:\/\/arxiv.org\/abs\/2607.13304\"\r\n        },\r\n        {\r\n          \"@type\": \"ScholarlyArticle\",\r\n          \"name\": \"Global is Good, Local is Bad?: Understanding Brand Bias in LLMs\",\r\n          \"url\": \"https:\/\/aclanthology.org\/2024.emnlp-main.707\"\r\n        },\r\n        {\r\n          \"@type\": \"Article\",\r\n          \"name\": \"The GCC AI Discovery Gap\",\r\n          \"url\": \"https:\/\/martawarren.com\/gcc-ai-discovery-gap\/\"\r\n        },\r\n        {\r\n          \"@id\": \"https:\/\/www.voctos.com\/research\/chatgpt-local-recommendations-egypt\/#article\"\r\n        }\r\n      ],\r\n      \"hasPart\": {\r\n        \"@type\": \"Dataset\",\r\n        \"@id\": \"https:\/\/www.voctos.com\/research\/chatgpt-appliance-brands-egypt\/#dataset\",\r\n        \"name\": \"ChatGPT and Google AI Mode appliance brand recommendations in Egypt, October 2026\",\r\n        \"description\": \"380 answers to Egyptian home appliance questions captured 13 to 14 October 2026: 60 Egyptian Arabic questions on logged-out ChatGPT, the same 60 again in Egyptian Arabic, in English and in Modern Standard Arabic, the same 60 on Google AI Mode signed out, and 40 search-style prompts on each engine. Coded for every brand recommended, mentioned or cautioned and every source domain, plus a same-day web-presence audit of 25 brands.\",\r\n        \"license\": \"https:\/\/creativecommons.org\/licenses\/by\/4.0\/\",\r\n        \"isAccessibleForFree\": true,\r\n        \"creator\": {\r\n          \"@id\": \"https:\/\/www.voctos.com\/#organization\"\r\n        },\r\n        \"temporalCoverage\": \"2026-10-13\/2026-10-14\",\r\n        \"spatialCoverage\": {\r\n          \"@type\": \"Place\",\r\n          \"name\": \"Egypt\"\r\n        },\r\n        \"measurementTechnique\": \"Pre-registered question bank and brand panel, with two further waves each registered before capture; logged-out browser capture, a new chat or fresh search per question, from a connection in Egypt; hand coding of recommended, mentioned and cautioned brands by blind coders; intervals resampling whole questions; generalizability study of brand x question x run; same-day audit of brand websites; Semrush Egypt search volumes added after the first run\",\r\n        \"variableMeasured\": [\r\n          {\r\n            \"@type\": \"PropertyValue\",\r\n            \"name\": \"Answers recommending Fresh\",\r\n            \"value\": \"31\",\r\n            \"unitText\": \"of 60\"\r\n          },\r\n          {\r\n            \"@type\": \"PropertyValue\",\r\n            \"name\": \"Recommendation slots going to made-in-Egypt brands\",\r\n            \"value\": \"29.8\",\r\n            \"unitText\": \"percent\"\r\n          },\r\n          {\r\n            \"@type\": \"PropertyValue\",\r\n            \"name\": \"Recommendations with the brand's own site cited\",\r\n            \"value\": \"20.2\",\r\n            \"unitText\": \"percent\"\r\n          },\r\n          {\r\n            \"@type\": \"PropertyValue\",\r\n            \"name\": \"Spearman correlation, search demand and recommendations, 25 brands\",\r\n            \"value\": \"0.914\"\r\n          },\r\n          {\r\n            \"@type\": \"PropertyValue\",\r\n            \"name\": \"Made-in-Egypt share of recommendation slots, English questions\",\r\n            \"value\": \"24.7\",\r\n            \"unitText\": \"percent\"\r\n          },\r\n          {\r\n            \"@type\": \"PropertyValue\",\r\n            \"name\": \"Repeated Arabic questions with the same first pick\",\r\n            \"value\": \"24\",\r\n            \"unitText\": \"of 47\"\r\n          },\r\n          {\r\n            \"@type\": \"PropertyValue\",\r\n            \"name\": \"Google AI Mode answers citing Facebook\",\r\n            \"value\": \"39\",\r\n            \"unitText\": \"of 60\"\r\n          },\r\n          {\r\n            \"@type\": \"PropertyValue\",\r\n            \"name\": \"Search-style ChatGPT prompts with a sponsored product unit\",\r\n            \"value\": \"24\",\r\n            \"unitText\": \"of 40\"\r\n          }\r\n        ],\r\n        \"distribution\": [\r\n          {\r\n            \"@type\": \"DataDownload\",\r\n            \"name\": \"Coded answers\",\r\n            \"encodingFormat\": \"text\/csv\",\r\n            \"contentUrl\": \"https:\/\/www.voctos.com\/wp-content\/uploads\/2026\/10\/voctos-egypt-appliances-chatgpt-coded-2026-10-v2.csv\"\r\n          },\r\n          {\r\n            \"@type\": \"DataDownload\",\r\n            \"name\": \"Brand audit table\",\r\n            \"encodingFormat\": \"text\/csv\",\r\n            \"contentUrl\": \"https:\/\/www.voctos.com\/wp-content\/uploads\/2026\/10\/voctos-egypt-appliances-brand-audit-2026-10.csv\"\r\n          },\r\n          {\r\n            \"@type\": \"DataDownload\",\r\n            \"name\": \"Raw answers\",\r\n            \"encodingFormat\": \"text\/csv\",\r\n            \"contentUrl\": \"https:\/\/www.voctos.com\/wp-content\/uploads\/2026\/10\/voctos-egypt-appliances-chatgpt-answers-2026-10-v2.csv\"\r\n          }\r\n        ]\r\n      }\r\n    },\r\n    {\r\n      \"@type\": \"BreadcrumbList\",\r\n      \"@id\": \"https:\/\/www.voctos.com\/research\/chatgpt-appliance-brands-egypt\/#breadcrumb\",\r\n      \"itemListElement\": [\r\n        {\r\n          \"@type\": \"ListItem\",\r\n          \"position\": 1,\r\n          \"name\": \"Home\",\r\n          \"item\": \"https:\/\/www.voctos.com\/\"\r\n        },\r\n        {\r\n          \"@type\": \"ListItem\",\r\n          \"position\": 2,\r\n          \"name\": \"Research\",\r\n          \"item\": \"https:\/\/www.voctos.com\/research\/\"\r\n        },\r\n        {\r\n          \"@type\": \"ListItem\",\r\n          \"position\": 3,\r\n          \"name\": \"Which appliance brands ChatGPT and Google AI Mode recommend in Egypt\",\r\n          \"item\": \"https:\/\/www.voctos.com\/research\/chatgpt-appliance-brands-egypt\/\"\r\n        }\r\n      ]\r\n    },\r\n    {\r\n      \"@type\": \"FAQPage\",\r\n      \"@id\": \"https:\/\/www.voctos.com\/research\/chatgpt-appliance-brands-egypt\/#faq\",\r\n      \"isPartOf\": {\r\n        \"@id\": \"https:\/\/www.voctos.com\/research\/chatgpt-appliance-brands-egypt\/#article\"\r\n      },\r\n      \"mainEntity\": [\r\n        {\r\n          \"@type\": \"Question\",\r\n          \"name\": \"Which appliance brand does ChatGPT recommend most in Egypt?\",\r\n          \"acceptedAnswer\": {\r\n            \"@type\": \"Answer\",\r\n            \"text\": \"Fresh. In 60 Egyptian Arabic appliance questions asked on 13 October 2026, ChatGPT recommended Fresh in 31 answers and in 9 of the 10 categories. Asked again it recommended Fresh in 28, in English in 27 and in Modern Standard Arabic in 35. Toshiba, LG and Beko follow it.\"\r\n          }\r\n        },\r\n        {\r\n          \"@type\": \"Question\",\r\n          \"name\": \"Does Google AI Mode recommend the same appliance brands as ChatGPT in Egypt?\",\r\n          \"acceptedAnswer\": {\r\n            \"@type\": \"Answer\",\r\n            \"text\": \"The same leaders in a different order. AI Mode recommended Tornado in 31 of 60 answers and Fresh in 29, and named 49 brands in all. It put an Egyptian brand first less often than ChatGPT, in 9 of 49 answers, and cited Facebook posts or groups in 39 of 60.\"\r\n          }\r\n        },\r\n        {\r\n          \"@type\": \"Question\",\r\n          \"name\": \"Why doesn't ChatGPT recommend my appliance brand in Egypt?\",\r\n          \"acceptedAnswer\": {\r\n            \"@type\": \"Answer\",\r\n            \"text\": \"In our data the strongest pattern was search demand. The five brands ChatGPT never named in 120 answers to the full Egyptian Arabic questions, Electrolux, Whirlpool, Gree, Black+Decker and Kenwood, are among the six least searched in Egypt, and four have no Egyptian website. Robots.txt rules and Wikipedia did not separate visible brands from missing ones.\"\r\n          }\r\n        },\r\n        {\r\n          \"@type\": \"Question\",\r\n          \"name\": \"Does ChatGPT prefer Egyptian or imported appliance brands?\",\r\n          \"acceptedAnswer\": {\r\n            \"@type\": \"Answer\",\r\n            \"text\": \"Neither overall, in Arabic. Egyptian brands took 29.8% of recommendations in the first run and 33.0% in the repeat, but 24.7% when the same questions were asked in English. They were the first pick for \\\"best brand\\\" in only 2 of 20 Arabic answers, but came first in 4 of 10 \\\"economy\\\" prompts on ChatGPT and 7 of 10 on AI Mode.\"\r\n          }\r\n        },\r\n        {\r\n          \"@type\": \"Question\",\r\n          \"name\": \"Does ChatGPT show ads when you ask about appliances in Egypt?\",\r\n          \"acceptedAnswer\": {\r\n            \"@type\": \"Answer\",\r\n            \"text\": \"Rarely for full questions, often for short ones. Sponsored product units appeared under 24 of 40 search-style prompts such as \u00ab\u0627\u0641\u0636\u0644 \u062a\u0643\u064a\u064a\u0641 \u0627\u0646\u0641\u0631\u062a\u0631 \u0641\u064a \u0645\u0635\u0631\u00bb, against 3 of 60 full questions in Modern Standard Arabic asked just before them.\"\r\n          }\r\n        },\r\n        {\r\n          \"@type\": \"Question\",\r\n          \"name\": \"Does ChatGPT recommend the same brands every time?\",\r\n          \"acceptedAnswer\": {\r\n            \"@type\": \"Answer\",\r\n            \"text\": \"Not exactly. Asked the same Arabic question twice, it kept the same first pick in 24 of 47 pairs and the identical brand list in 11 of 54. But the leading brands kept appearing: Fresh was recommended in 31 and then 28 of 60 answers, LG in 18 and 16.\"\r\n          }\r\n        },\r\n        {\r\n          \"@type\": \"Question\",\r\n          \"name\": \"Can I trust the service hotlines ChatGPT gives for appliance brands in Egypt?\",\r\n          \"acceptedAnswer\": {\r\n            \"@type\": \"Answer\",\r\n            \"text\": \"In our data, the hotlines ChatGPT wrote into its answers were the brands' own, such as Toshiba Egypt's 19111. The sources behind them were another matter: in 3 of 10 service answers ChatGPT cited third-party sites that use a brand's name, one of them advertising a different Toshiba number. Check any number against the brand's own site before you call.\"\r\n          }\r\n        }\r\n      ]\r\n    }\r\n  ]\r\n}\r\n<\/script>\r\n\r\n","protected":false},"excerpt":{"rendered":"<p>Fresh was recommended in 31, 28 and 27 of 60 answers across two Arabic runs and one English run. ChatGPT&#8217;s appliance picks track Egyptian search demand; five brands never appeared in 120 Arabic answers.<\/p>","protected":false},"author":13,"featured_media":0,"template":"","meta":{"_acf_changed":false,"inline_featured_image":false,"footnotes":""},"categories":[],"tags":[],"research_type":[],"class_list":["post-14223","research","type-research","status-publish","hentry"],"acf":[],"_links":{"self":[{"href":"https:\/\/www.voctos.com\/ar\/wp-json\/wp\/v2\/research\/14223","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.voctos.com\/ar\/wp-json\/wp\/v2\/research"}],"about":[{"href":"https:\/\/www.voctos.com\/ar\/wp-json\/wp\/v2\/types\/research"}],"author":[{"embeddable":true,"href":"https:\/\/www.voctos.com\/ar\/wp-json\/wp\/v2\/users\/13"}],"version-history":[{"count":3,"href":"https:\/\/www.voctos.com\/ar\/wp-json\/wp\/v2\/research\/14223\/revisions"}],"predecessor-version":[{"id":14231,"href":"https:\/\/www.voctos.com\/ar\/wp-json\/wp\/v2\/research\/14223\/revisions\/14231"}],"wp:attachment":[{"href":"https:\/\/www.voctos.com\/ar\/wp-json\/wp\/v2\/media?parent=14223"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.voctos.com\/ar\/wp-json\/wp\/v2\/categories?post=14223"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.voctos.com\/ar\/wp-json\/wp\/v2\/tags?post=14223"},{"taxonomy":"research_type","embeddable":true,"href":"https:\/\/www.voctos.com\/ar\/wp-json\/wp\/v2\/research_type?post=14223"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}