ChatGPT Ads Now in 89 Countries: Is It Worth Testing?

If ChatGPT ads have just switched on in your country, you’re probably asking whether they deserve a slice of next month’s budget. On 21 October 2026, OpenAI added 26 more countries, bringing the total to 89. So the list of ChatGPT ads countries now covers most of the Americas and Europe and a growing share of Africa and Asia. My short answer: a small, well-measured test is worth running if your customers research purchases in conversation and you can track conversions properly. It isn’t worth it if you can’t measure results, or you sell in a category where ads won’t show.

In this guide I cover what was confirmed this week, how the ad system picks and prices ads, who it suits, what to spend, how to write ads for it, how to measure it, and a 30-day plan you can copy.

What OpenAI Announced on 21 October

OpenAI made three announcements on the same day. Together, they give the clearest picture yet of where the product stands.

• 26 new countries, 89 in total. Fourteen are in Latin America. Twelve are spread across Europe, the Middle East, Africa and Asia.
• More than 100,000 advertisers have run ads on ChatGPT since the launch just over eight months ago. That is a total over time, not the number of advertisers active today.
• A technical write-up of the ad system, explaining how ads are chosen, ranked and priced, and how ad choice is kept apart from writing answers.

This follows a busy few months. Ads began in the US on 9 February. Brazil and Mexico went live in mid-August. 31 European markets followed from 24 August, and India launched on 29 August. Seven Southeast Asian markets and Taiwan went live on 23 September. On 31 August, about 200 days after launch, OpenAI said the ad business had passed a $1 billion annualized revenue run rate.

Keep that $1 billion figure in proportion. A run rate takes recent revenue and projects it across a full year. It tells you advertisers are spending. It says nothing about whether any single advertiser is getting a return. I always keep those two things apart.

Which Countries Were Added

OpenAI named 11 of the 26 new markets:

• Latin America: Argentina, Chile, Colombia and Peru (four of the fourteen).
• Africa: Nigeria and South Africa.
• Europe: Serbia and Ukraine.
• Asia: Cambodia, Bangladesh and Pakistan.

The other 15 weren’t named in the announcement, and no start date was given for single countries. Don’t plan around a list from a third party. Log in to Ads Manager and check the availability list and the targeting options your own account offers.

Who Can Buy, and Who Sees the Ads

In the new markets you can buy through OpenAI’s technology partners, or yourself through Ads Manager. Two points are easy to miss.

• Your billing entity matters. OpenAI’s help centre ties self-serve access to the country of the legal entity that is billed. That country has to be on the available list. A country being on the list of ad markets doesn’t mean every business there can self-serve on day one. India, for example, launched in August with buying only through OpenAI’s sales team and partners.
• Only part of the audience sees ads. Ads appear for logged-in users on the Free and Go plans. Plus, Pro and Enterprise users don’t see them. When it announced ads in January, OpenAI also said they wouldn’t be shown to users under 18 or users predicted to be under 18. They also wouldn’t appear next to sensitive topics like health, mental health and politics.

OpenAI doesn’t publish how many people in each country are on Free or Go. So you can’t size your audience in advance. You’ll only learn it from impressions once you launch.

For background on how the platform opened up, see our explainer on ChatGPT Ads Manager going self-serve and our coverage of the European launch.

How ChatGPT Picks and Prices an Ad

I’d read the technical write-up before you spend anything. It explains why ChatGPT ads don’t behave like search ads.

Two Stages: Nomination, Then Auction

First, a custom, specially trained language model that OpenAI calls the orchestrator reads the conversation so far. It writes intent queries, pulls candidate ads from an index that matches by meaning, and nominates finalists. OpenAI says the orchestrator doesn’t decide which ad is served and doesn’t set the price.

A separate auction then handles admission, pacing, bid calculation, quality-adjusted ranking, pricing and the final pick. Candidates are scored on relevance, predicted engagement and past performance, as well as the bid.

In practice, you can’t simply outbid a more relevant ad. A clear, specific ad that people engage with should win more auctions at the same bid than a vague one.

Context, Not Keywords

You don’t bid on search terms. Matching reads intent as it builds up over several turns of a conversation, not just one question. You guide it with plain-language context hints at the ad group level. Your ad copy and landing page also feed into the match.

If a user has turned on ad personalization, the system can also use non-sensitive details from their past conversations. If they’ve turned it off, only the current conversation counts.

Why Your Ad Sometimes Won’t Show

Several things can stop an eligible ad from appearing:

• The orchestrator decides it doesn’t fit. It can block an ad that passed policy review but would feel out of place in a certain conversation.
• The ad is too slow. If ChatGPT’s answer is ready before the ad, the ad request times out and no ad is shown. OpenAI says it doesn’t delay answers to wait for ads, and treats showing no ad as a normal result.
• The conversation is not eligible. That covers the excluded topics and users described above.

Expect lower and more uneven delivery than on a search platform, especially in a market that has only just opened.

What Stays Separate

Ad selection runs as its own service, with separate models and code. OpenAI says ChatGPT’s answers don’t depend on the ads system. It also says advertisers don’t see conversations and it doesn’t sell user data.

For you, that means two things. Buying ads won’t get your brand recommended inside the answer. And you won’t get a search-terms report, so measurement has to come from your own tracking.

Are ChatGPT Ads Worth Testing? Who They Suit

The format rewards businesses whose customers think out loud before they buy. Here’s the rough fit check I’d use.

Good Candidates

• Considered purchases. Software, travel, home improvement, education, B2B services and higher-priced retail, where people compare options and ask follow-up questions.
• Retailers with clean product data. If you already run shopping campaigns elsewhere, product feed campaigns in ChatGPT ads let you reuse much of that work.
• Brands already selling in the newly opened markets. If you have local landing pages, local pricing and a billing entity in one of the new countries, you can test without building anything from scratch.
• Teams with solid conversion tracking. Pixel or server-side conversions, consistent UTMs and an analytics setup that separates paid from organic AI traffic.

Poor Candidates, for Now

• Health, mental health and political advertisers. The conversations you’d most want to appear in are excluded.
• Businesses that can’t track conversions. Without tracking, you’re judging on clicks alone. I wouldn’t start a test in that position.
• Very small budgets already stretched across channels. If you can’t commit the minimum daily spend for a month, the data won’t tell you much.
• Audiences mostly on paid ChatGPT plans. If your buyers are professionals on company-paid Enterprise accounts, many of them won’t see ads.

Costs, Bidding and Budgets as Documented

OpenAI hasn’t published average costs or benchmarks for any market. Treat any CPC or CPM figure quoted elsewhere as someone else’s experience, not a price list.

Here is what is documented:

• Two bid models. CPM, where you pay per thousand impressions, and CPC, where you pay per click. Our guide to ChatGPT ads bidding covers the cost controls in more detail. CPC arrived with the self-serve beta in May.
• Minimum daily spend by currency. The help-centre table published in August set floors like US$25, £15, CA$25, A$25, ¥2,500, ₩25,000, R$40, MX$150 and ₹725 a day. If your market is one of the new 26, check the table in Ads Manager for your billing currency. Don’t assume US dollar terms.
• Budgets and pacing. You set campaign budgets and bids, and the auction spreads spend across the day.

For a first test, I’d usually pick CPC, because you pay only for visits. Switch to a CPM test only once you know your click-through and conversion rates. Then you can work out whether paying for impressions is cheaper per conversion.

A simple way to set your test budget is to work backwards from conversions. Say you want around 30 conversions to judge results, and you expect a 3% conversion rate from your landing page. Then you need about 1,000 clicks. Your actual CPC sets the cost. If that total is more than you can risk, narrow the test to one product or service so the conversions you get can be compared.

Ads That Fit a Conversation

A ChatGPT ad appears below the answer and is labelled as sponsored. It’s built from a few short parts: your business name and icon, a headline, a short description, an image and a link. Product feed campaigns pull titles, images, prices and ratings from your feed instead.

Because the ad follows a helpful answer, it needs to read as the next useful step, not an interruption.

• Continue the conversation. If someone is planning a week of family meals, “Get these recipes delivered with portioned ingredients” fits better than “Award-winning meal kits.”
• One benefit per ad. There isn’t room for a feature list, and relevance scoring favours a clear match.
• Write context hints as situations. “A small agency owner comparing project tools for a team of ten” gives the orchestrator far more to work with than “project management software.” In my experience, this is where most new advertisers write hints that are too short.
• Match the landing page. The page feeds into matching. So send traffic to a focused page that uses the same words as your hints. A generic homepage weakens both relevance and conversion.
• Localize properly. In new markets like Argentina or Colombia, write native Spanish copy and hints and use local prices. Translated US copy will likely score worse on relevance.

OpenAI says it collects clicks, hides and survey feedback from users. An ad people hide is likely to hurt you in quality-adjusted ranking, so avoid clickbait headlines.

Measurement: Prove It or Stop

Ads Manager reports impressions, clicks, spend, click-through rate, average CPC and CPM, and conversions. Conversions come from the OpenAI pixel or the Conversions API, and they’re reported as totals. You won’t see the conversations behind them.

Set up three layers before you spend anything. I treat this as non-negotiable:

1. Platform conversions. Install the pixel or connect the Conversions API, and define the action that matters, like a purchase, lead or booked demo.
2. Clean UTMs. Tag every destination URL with a paid medium, for example utm_source=chatgpt&utm_medium=paid&utm_campaign=co-launch-oct. Organic ChatGPT citation links often carry utm_source=chatgpt.com, so a separate medium keeps the two apart. Our UTM naming system covers how to keep tags consistent.
3. Analytics that catches the rest. Some ChatGPT visits arrive without a referrer and land in Direct. Fix that first with our guide to ChatGPT traffic showing as direct in GA4.

Judge the test on cost per acquisition and conversion quality, compared with your search and social campaigns over the same period. Click-through rate on its own tells you little.

Risks to Weigh Before Testing ChatGPT Ads in New Countries

• Thin delivery. Excluded users and topics, partial reach and ad timeouts can mean fewer impressions than you’d expect from your budget.
• Limited transparency. No query or conversation data means slower improvement than on search platforms.
• A moving product. OpenAI says more formats, buying options and measurement tools are coming, with no dates. Screens and policies may change mid-test.
• Brand safety. The orchestrator screens for fit, but you still can’t see where your ad appeared. Brand-sensitive advertisers should start with narrow context hints.
• Market gaps. In some new countries, self-serve buying, local billing or partner coverage may not be ready on day one. Confirm before you build a plan around it.
• Opportunity cost. Every pound or peso here comes out of a channel you already understand. Cap the test and set a decision date.

A 30-day ChatGPT Ads Test Plan

This plan suits a single business or one client testing one offer in one country. It’s how I’d structure a first test.

Days 1–5: Set Up and Check

1. Confirm your country and billing entity are eligible for self-serve, or contact a partner.
2. Register in Ads Manager and add billing in your local currency.
3. Install the pixel or Conversions API and fire a test conversion.
4. Build UTMs and confirm paid ChatGPT visits appear as a separate source in analytics.
5. Write a decision rule now, for example: “Continue if CPA is within 30% of our search CPA by day 30.”

Days 6–7: Build and Launch

1. Create one campaign on CPC at or a little above the minimum daily spend.
2. Add two ad groups for the same offer: one with narrow, situation-based context hints and one with broader topic hints.
3. Write two ads per ad group, one benefit-led and one problem-led, all pointing to the same focused landing page.
4. Submit for review and launch once approved.

Days 8–21: Learn Without Fiddling

1. Check delivery daily. If an ad group gets almost no impressions after a few days, broaden or rewrite its hints.
2. Change one thing per week. Leave bids and budget alone while you gather data.
3. Note which landing pages, products and times of day bring conversions.

Days 22–30: Decide

1. Compare CPA, conversion rate and lead or order quality against search and social for the same period.
2. Pause the weaker ad group and ad. If the stronger one meets your rule, raise its budget slowly.
3. If results are borderline, try one CPM version before you decide. If they’re clearly weak, stop and write down what you learned for next time.

My Verdict on ChatGPT Ads in 89 Countries

With 89 ChatGPT ads countries, more than 100,000 advertisers and a published account of how ads are chosen, this is now a real channel, not an experiment for big brands only. It is still a young one. Costs aren’t benchmarked, delivery is partial and reporting only shows totals.

If your customers research before they buy and you can track conversions, run the 30-day test with a fixed budget and a written decision rule. Your next step is measurement. Before you launch, fire a test conversion and confirm paid ChatGPT visits appear as their own source in analytics. If that works, the rest of the test is just discipline.