Context hints are the only targeting lever ChatGPT ads give you. There are no keywords, no match types and no audience interests to fall back on. You describe the conversation you want to appear in, and OpenAI decides whether your ad fits it.<\/p>\n\n\n\n
In one published nine-day test, the same hints produced a click-through rate 6.5 times higher in the UK than in New Zealand. How well a hint describes real conversations decides most of what you pay.<\/p>\n\n\n\n
Context hints are short, plain-language descriptions you add to each ad group in ChatGPT Ads Manager. They tell OpenAI which conversations your offer suits. OpenAI’s help center describes them as broad thematic signals that guide matching, not exact-match keywords, and says they do not guarantee delivery in any specific conversation.<\/p>\n\n\n\n
The matching system reads your hints alongside the ad title, the ad copy and the landing page, then makes the final relevance call. That makes a hint closer to a brief than a bid.<\/p>\n\n\n\n
A Google Ads keyword names what someone typed. A context hint names a person in a situation. “Dental implants Riyadh” is a keyword. “Adults in Riyadh comparing the cost of dental implants with bridges after losing a back tooth” is a hint.<\/p>\n\n\n\n
A keyword is matched as a string. A hint is interpreted as meaning across the whole conversation, so one well-written hint covers dozens of ways of asking the same thing. You stop listing variants and start describing the buyer.<\/p>\n\n\n\n The last row matters most. Without a search terms report, you cannot see which conversations triggered your ad. The hint is both your targeting and your only diagnostic, so a vague hint leaves you guessing about everything downstream.<\/p>\n\n\n\n A strong hint names four things: who the person is, what they are trying to solve, the constraint that shapes the decision, and where they are in the decision. Leave out any one and the matcher has to guess, usually in the direction of more impressions and fewer buyers.<\/p>\n\n\n\n A simple pattern holds most good hints together:<\/p>\n\n\n\n Keep each hint to one or two sentences. If a hint grows longer, it usually describes two buyers, and those belong in separate ad groups. OpenAI’s own guidance gives the same rule: meaningfully different products, audiences or use cases go in different ad groups.<\/p>\n\n\n\n A keyword names what someone typed. A context hint names a person in a situation.<\/p><\/figure>\n\n\n\n Good hints read like a sentence from a real chat in your market, with the local constraint left in. The examples below are drafts we would start from, not tested results. Each pairs a weak, keyword-style hint with a stronger rewrite.<\/p>\n\n\n\n Notice what the stronger versions keep: a city, a constraint and a decision. Those are the details a person types into ChatGPT and leaves out of a Google search.<\/p>\n\n\n\n Write hints in the language your buyers chat in, and test both when your market is bilingual. OpenAI matches on meaning, so an English hint can still describe an Arabic conversation. Nobody has yet published Arabic-language ChatGPT ads results, so treat language as a variable to measure, not a settled rule.<\/p>\n\n\n\n In practice, split by language at the ad group level, the same way you would split Arabic and English search campaigns. That keeps the ad copy and landing page consistent with the conversation, and the matcher weighs all three together.<\/p>\n\n\n\n Two Gulf-specific details deserve a place in the hint when they apply. Name the emirate or city if delivery, licensing or price differs by location. Name the season or occasion, such as Ramadan, Eid, back-to-school or summer, because those change what people ask and when.<\/p>\n\n\n\n Rejections so far point to one pattern: ads that read as if the AI itself is making a promise. Synter, a paid media software company, had three ads flagged under OpenAI’s “platform integrity and misrepresentation” reason in June 2026. All three claimed their product acts autonomously, and the rewrites that passed led with what the user gets.<\/p>\n\n\n\n That lesson travels well beyond software. Inside a ChatGPT answer, a card that sounds like the assistant vouching for you is the risk. Keep a human in the sentence, describe the outcome plainly and avoid claims the landing page cannot back up.<\/p>\n\n\n\n Regulated categories need extra care in the Gulf. Health, finance and real estate claims face local advertising rules on top of OpenAI’s policies. A hint can describe the need, but the ad and landing page still have to meet both sets of rules.<\/p>\n\n\n\n The hint gets you into the conversation. The ad copy has to survive it. Anicca Digital’s July 2026 UK trial found that ad text is truncated heavily and the image renders small, so the offer has to sit in the first few words and the image works best as a brand cue.<\/p>\n\n\n\n Write the title as the answer to the situation in your hint. If the hint describes someone comparing implants with bridges, the title should speak to that comparison, not to “Riyadh’s leading dental clinic”. The landing page should then continue the same thread, because OpenAI weighs it in relevance as well.<\/p>\n\n\n\n Keep one message per ad group. The chat card has a title, one or two sentences of body text, an image and a link. There is no room for the stacked headlines you would test in Google responsive search ads.<\/p>\n\n\n\n Read click-through rate per ad group and per country in the first week, then look past the click. Ads Manager reports impressions, clicks, spend, click-through rate, cost per click and conversions. It does not show which conversations matched, so you judge a hint by what it delivers.<\/p>\n\n\n\n<\/th> Google Ads keyword<\/th> ChatGPT context hint<\/th><\/tr><\/thead> What it targets<\/td> A search query<\/td> A person in a specific situation<\/td><\/tr> How it matches<\/td> Exact, phrase or broad string match<\/td> Meaning across the whole chat<\/td><\/tr> Coverage<\/td> One phrasing per keyword<\/td> Many phrasings of one need<\/td><\/tr> Negative control<\/td> Negative keywords<\/td> Exclusions written into the hint<\/td><\/tr> Delivery<\/td> Eligible when the query matches<\/td> No guarantee, the system decides relevance<\/td><\/tr> Reporting<\/td> Search terms report<\/td> No search terms report yet<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n How do you write a context hint that works?<\/h2>\n\n\n\n
\n
What do good hints look like for Gulf and Egyptian businesses?<\/h2>\n\n\n\n
Business<\/th> Weak hint<\/th> Stronger hint<\/th><\/tr><\/thead> Dental clinic, Riyadh<\/td> Dental implants Riyadh<\/td> Adults in Riyadh comparing dental implants with bridges after losing a back tooth, who want to know total cost and recovery time before booking a consultation<\/td><\/tr> Online fashion store, UAE<\/td> Abayas online UAE<\/td> Women in the UAE looking for lightweight everyday abayas for summer heat, comparing fabric and delivery times before Eid<\/td><\/tr> B2B software, Saudi Arabia<\/td> ERP software KSA<\/td> Finance managers at Saudi companies of 50 to 300 staff who need ZATCA-compliant e-invoicing and are comparing ERP systems before the next integration deadline<\/td><\/tr> Language school, Cairo<\/td> English courses Cairo<\/td> Working professionals in Cairo preparing for IELTS within three months who need evening or online classes<\/td><\/tr> Property developer, Dubai<\/td> Apartments for sale Dubai<\/td> Expat families in Dubai comparing off-plan and ready apartments near schools, deciding whether to buy instead of renewing a lease<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n Should you write context hints in Arabic or English?<\/h2>\n\n\n\n
Why do some ads get rejected, and how do you avoid it?<\/h2>\n\n\n\n
How should the ad copy work with the hint?<\/h2>\n\n\n\n
How do you know whether a hint is working?<\/h2>\n\n\n\n