What ChatGPT Ads Cost in the Arab World: OpenAI Publishes a Floor, Not a Price

OpenAI has never published a price for ChatGPT ads. It publishes a floor, and in almost every Arab market that floor is 25 US dollars a day, which is the only official cost figure that exists.

Everything else you have read is an estimate. The 25 dollar CPMs, the 60 dollar default bids, the 3 to 5 dollar clicks: those come from agencies, not from OpenAI.

So we read the primary sources instead. OpenAI’s own documentation, its policy changelog, and the location catalogue it publishes as a raw CSV file.

We found three things that have not been reported anywhere, and one of them changes what a Gulf advertiser should do this month. For which countries can buy at all, start with our earlier report on the eleven Arab countries where ChatGPT Ads opened.

Key takeaways

• OpenAI states in its own FAQ that ChatGPT Ads has no performance benchmarks yet, and in its conversion campaign guide that there is no recommended bid amount.
• The only official price is a minimum daily budget by billing currency. It is 25 USD, and of the eleven available Arab markets only Qatar has its own currency on the list.
• The United Arab Emirates is the single country in the world that OpenAI lets you target but not buy from. AED even has a published minimum.
• Saudi Arabia appears in none of OpenAI’s three published files. Not as an advertiser country, not as a currency, not as a targetable location.
• Governorate-level targeting is already live across the region. OpenAI’s catalogue lists 188 sub-country regions across the eleven open Arab markets, including 27 Egyptian governorates.
• Health, financial and legal services outside the United States are still prohibited, so for a large part of the region’s paid search economy the price is not the blocker.

What does OpenAI actually publish about price?

Two numbers, and neither is a benchmark. OpenAI publishes a minimum daily budget per billing currency and a bidding structure, and that is the whole of it. Nowhere does it publish an expected cost per click, an expected CPM, or a recommended bid for any market or any category.

That is not an omission we inferred from silence, because OpenAI writes it twice in its own words, in two separate documents.

The ChatGPT Ads FAQ says the platform “does not yet have performance benchmarks across advertisers, industries, or campaign types.”

The conversion campaign guide is blunter still. Buried in a list of optimization tips, between advice on tracking hygiene and advice on creative volume, sits the sentence: “There is no recommended bid amount at this time.”

So when a vendor quotes you a ChatGPT CPM for Cairo or Doha, ask which OpenAI page it came from.

There is no such page, so the number came from somewhere else. Either it is their own account data, which is a sample of one, or it is an estimate presented as a rate card.

There are three ways to buy. CPM bills per thousand impressions and optimises for reach, while CPC bills per valid click and optimises for traffic.

oCPC optimises toward a conversion event you define, but it still bills per click rather than per conversion.

You cannot convert an existing campaign from one objective to another, so the objective you choose at creation is the objective you live with for that campaign’s whole life.

What is the real minimum to run ChatGPT ads in an Arab market?

Twenty-five US dollars a day, for almost every advertiser in the region. OpenAI publishes a minimum daily budget table covering 21 billing currencies, and of the eleven Arab countries with self-service access, exactly one has its own currency on that list.

Qatar is on that list, at 70 QAR. So is AED, at 70 dirhams, and we will come back to why a currency appears for a country that cannot open an account.

There is no EGP, no KWD, no BHD, no OMR, no JOD, no LBP, no IQD, no MAD, no TND and no DZD.

That leaves ten of the eleven markets billing in dollars, so an Egyptian, Jordanian, Kuwaiti, Bahraini, Omani, Lebanese, Iraqi, Moroccan, Tunisian or Algerian advertiser budgets in US dollars and carries the currency risk on every invoice.

Confirm your billing currency during signup before you commit. OpenAI states that billing country, currency and legal entity are tied to account setup and may not be editable afterwards.

Twenty-five dollars a day works out at roughly 750 dollars a month, per campaign, simply to keep that campaign delivering.

OpenAI’s own best practices recommend separate campaigns for separate objectives. Run three and your floor is roughly 2,250 dollars a month, before you have bought a single useful click.

That is the honest entry price, and it is worth being precise about what it means. It is the number below which the platform will not let your campaign run at all, which is a different thing from a rate card or a media plan.

Why is the UAE in OpenAI’s targeting file but not its advertiser list?

Because the billing and delivery rails are built while the commercial gate is not open yet. This was the finding we did not expect when we started reading, and anyone with a browser can check the whole thing in about four minutes.

OpenAI publishes three separate things: an Ads Manager availability list, a minimum daily budget table, and a geotargeting catalogue offered as a downloadable CSV.

The UAE is absent from the first list and present in the other two.

The catalogue contains 53 countries. The availability list contains 52. We compared them line by line, and the United Arab Emirates is the only genuine difference.

The CSV lists the UAE at ID 1000001, country code AE. All seven emirates appear under it as separate targetable regions: Abu Dhabi, Dubai, Sharjah, Ajman, Ras al Khaimah, Fujairah and Umm al Quwain.

The AED minimum daily budget of 70 dirhams sits in the same published table as USD and QAR, for a country that cannot yet open an account.

A UAE company still cannot open a self-service account, but an Egyptian or Qatari entity can target UAE users today. That asymmetry is live right now, and for a regional group with more than one legal entity it is worth acting on this month.

Saudi Arabia is a different picture entirely.

It is not in the availability list, SAR is not in the currency table, and Saudi Arabia does not appear anywhere in the geotargeting catalogue.

So you cannot advertise from Saudi Arabia, and you cannot target Saudi Arabia either. On the evidence OpenAI publishes today, the UAE looks considerably closer to opening than Saudi Arabia does.

We are not claiming to know OpenAI’s roadmap, and we would not trust anyone who says they do. What we are reporting is narrower and checkable: two of the three files a market needs already carry the UAE, and none of them carry Saudi Arabia.

Can you target an Egyptian governorate in ChatGPT Ads?

Yes, and OpenAI’s own help documentation does not tell you that anywhere.

The campaign setup guide mentions country targeting, then state, DMA and ZIP targeting “in the United States”. We read the catalogue instead of the prose, and the picture changes.

Across the eleven open Arab markets, OpenAI’s catalogue lists 188 sub-country regions that you can target today.

Algeria has 58 wilayas and Egypt has 27 governorates, which between them account for nearly half the total. Tunisia has 24, Iraq 18, Jordan 12, Morocco 12, Oman 11, Lebanon 8, Qatar 8, Kuwait 6 and Bahrain 4. The UAE adds 7 emirates on top, targetable by advertisers based elsewhere.

Region is the finest grain available anywhere outside the United States, and the catalogue makes that limit explicit.

The catalogue contains exactly five postal code entries. All five are American. We found no city targeting and no radius targeting anywhere in the file at all.

For a Cairo retailer working from a small budget, that is still a meaningful control.

You can concentrate a 25 dollar daily floor on Greater Cairo, Alexandria and Giza rather than spreading it across 27 governorates. That is the difference between a test that produces signal and a test that produces noise.

What will actually hit your bank account

Four mechanics decide what actually leaves your account, and none of the four is the bid you set.

This surprises advertisers coming from Google or Meta. The budget you set is neither the amount that leaves your account, nor the schedule on which it leaves.

The payment threshold, not the budget, triggers the charge. Every account gets an assigned threshold, and OpenAI’s own worked example uses 25 dollars. Your card is charged each time unpaid spend reaches it. Thresholds rise automatically with payment history, and OpenAI states plainly that support will not change one for you.

A daily budget is a seven-day average, not a daily cap. Spend can reach twice your daily budget on a single day. Over seven days it will not exceed seven times the daily figure. So a 25 dollar daily budget can take 50 dollars out on a Tuesday.

Adding a card triggers a temporary 100 dollar authorisation. It is a hold, not a charge. OpenAI says it may stay visible on your statement for up to seven days, which is long enough to alarm a finance team that was not warned.

Pausing a campaign does not stop billable delivery immediately. OpenAI states that ads may continue to appear for up to 24 hours after a pause, and that any spend incurred during that window remains billable to you, which is a detail we would want written into any internal approval process before a first test.

One more constraint catches regional groups in particular, and it is the kind of detail that stops a launch on the morning it was meant to go live.

Your payment method must match the country of the entity that opened the account. A card issued elsewhere may be rejected, and a different billing country means a separate advertiser account.

How does this compare with Google Ads?

On price transparency the two platforms are not comparable at all, and that is the honest headline rather than a diplomatic one.

Google publishes forecasting tools, historical cost data, and no minimum daily budget at all. OpenAI publishes a minimum and explicitly declines to publish benchmarks.

Here is the comparison we use with clients when the question is where the next 2,000 dollars should go, whether that is into paid search or somewhere else.

Two lines in that table decide most real cases.

The category line means that for a Cairo laboratory, a Doha bank or an Amman law firm, cost is not the deciding factor at all, because they are not eligible to buy at any price.

OpenAI’s Ad Policies, at version 1.5 dated 31 August 2026, still read: “Ads for financial services outside the US are generally prohibited.” The same sentence appears for health services. Legal services are permitted in the US only.

The intent line means the two platforms are not doing the same job, even when they are shown to the same person on the same question. If ChatGPT is where your buyers now ask the question, AI search optimization is the other half of the answer.

A Google keyword is a control. A ChatGPT context hint is a description, and OpenAI says plainly that hints are “not exact-match controls, audience targeting rules, or instructions to show your ad only for specific conversations.”

You describe when your product is useful, and a model decides. That is a genuinely different discipline from keyword buying, and the minimum daily budget means you learn it at 750 dollars a month rather than at 50.

My read on this

I think the interesting question is not what a ChatGPT click costs. It is what OpenAI is signalling by declining to publish a number at all, in a category where every other major platform publishes one.

Platforms publish benchmarks when they want volume, which is why Google publishes forecasts: forecasts pull advertisers in.

OpenAI does the opposite. It publishes a floor and no benchmark, and it defaults new ad groups to an automatic bid strategy that carries no CPA, CPC or ROAS guarantee of any kind.

My reading is that this is a platform recruiting fewer, larger and more patient advertisers during a beta, rather than a long tail of 5 dollar-a-day tests. I could be wrong about the motive, though the behaviour itself is documented on their own pages.

The 25 dollar minimum does the same job from the other direction. It is not a large sum, and I do not think it is meant to be. It is meant to filter.

For most advertisers in the eleven open markets, my recommendation is unchanged from what we wrote when those countries opened.

If you are eligible and your category is allowed, run a real test at the floor for a full month across a single region rather than a week across a whole country. A week of delivery at the minimum will not give you enough data to read.

If you are not eligible, the organic answer above the ad still reaches every plan, every tier and every country, Saudi Arabia and the UAE included, because it has no country gate on it. That is what ChatGPT optimization is for.

What this report does not tell you

We can be specific about three limits on what this report covers.

It does not tell you what you will actually pay per click or per thousand impressions in your own market.

Nobody can tell you that yet, and that includes OpenAI, so any figure you are quoted is an estimate drawn from somebody else’s account.

It does not predict when Saudi Arabia or the UAE will open. We are reporting what appears in three published files today.

Rails being built is evidence of intent rather than a schedule, and OpenAI has given no date, so we are not going to guess one for them.

This report is accurate on its publication date and no longer.

OpenAI’s availability page moved from 9 countries to 52 in under three weeks, and the billing article we cite here had been updated 11 hours before we read it. Verify against OpenAI’s own pages before you act on anything in this report.

Objections we hear

“Our agency quoted us a ChatGPT CPM. Are they lying?”

Not necessarily, but whatever they are quoting from, it is not OpenAI.

The only ways to hold a CPM figure are your own delivery data or somebody else’s, so ask them which account it came from, over what period, and in which countries. A CPM from a US consumer-goods account in April tells you very little about a Tunisian one in September.

“25 dollars a day sounds cheap. Why is that a filter?”

Because the floor applies per campaign, and OpenAI recommends separate campaigns for separate objectives, regions and product categories.

A serious structure is three or four campaigns, which puts you at 2,250 to 3,000 dollars a month before you have optimised anything. The daily figure looks small on its own, and the structure it sits inside is what makes it expensive.

“We are a Dubai group. Can we advertise through our Egyptian entity?”

Legally and operationally that is a decision for you and your auditor rather than for us.

What we can give you is the mechanics. The entity that advertises and is billed must be based in a listed country. The payment method must match that entity’s country. The UAE is targetable from an eligible account today.

Get the entity question answered before anyone on your team touches the signup flow, because currency and billing country are effectively locked at account creation and OpenAI will not swap them for you afterwards.

“Should we wait for the price to come down?”

There is no published price for anyone to bring down in the first place.

The variable that actually moved this year is access, from 9 countries to 52 in a matter of weeks. The category policy has stayed where it was through five revisions since March 2026.

The more useful question

The question is not what ChatGPT ads cost. It is what a 750 dollar month actually buys you in learning while the platform still has no benchmarks.

If your category is allowed and your entity sits in one of the eleven markets, that month buys you real delivery data on a channel where nobody has benchmarks.

This is the only period in a platform’s life when a small advertiser and a large one are working from exactly the same information, and it does not last. That window closes the day OpenAI starts publishing benchmarks.

If your category is blocked, or your entity sits in Riyadh or Dubai, the same money buys something with no country gate and no category gate on it.

It buys you the chance to be the answer the model gives above the ad, which is the position no auction can take away from you.

We would put the first dirham there either way, and we have told clients in open markets the same thing. Ask us for an AI search audit and we will show you which answers you already appear in and which ones your competitors own.

Related guides

• ChatGPT Ads Just Opened in 11 Arab Countries. Saudi Arabia and the UAE Are Not Among Them
• ChatGPT Ads Aren’t Available in Any Arab Market Yet. Here’s What Actually Works Right Now
• Google Ads in 2026: What Every Egyptian Brand Should Know (And the AdWords History That Still Matters)
• Social Media Marketing for Egyptian Brands in 2026: A Senior, Channel-by-Channel Playbook