ChatGPT Ads Reach 89 Countries. Do They Actually Sell?

ChatGPT ads have scaled faster than almost any new ad channel, but the first independent evidence suggests they sell much less per visit than paid search. For most advertisers, the right move in late 2026 is a small, ring-fenced test with its own measurement, not a budget shift.

How big has the ChatGPT ad business become?

Very big, very quickly. According to The Next Web, OpenAI’s ad business reached an annualised $1 billion in August 2026, up from $100 million in March. On October 21, 2026 OpenAI added 26 markets, bringing ChatGPT ads to 89 countries, and said more than 100,000 advertisers had run campaigns since launch. That is a cumulative count of organisations that have tried the channel, not the number buying today.

Ads are shown to users on the Free and Go plans. Plus, Pro and Enterprise subscribers do not see them. OpenAI says ads are “clearly labeled and separate from ChatGPT’s answers”, and that conversations are kept private from advertisers. Advertisers can buy through partners or through OpenAI’s self-serve Ads Manager, and a format called Sponsored Agents began testing with some US advertisers in September.

What does the first independent study find?

In October 2026, Maximilian Kaiser (WU Vienna) and Christian Schulze (Frankfurt School of Finance & Management) published a working paper, still under peer review, using data from the market-intelligence firm Grips Intelligence. It covers almost 10 billion sessions between June 2025 and August 2026.

For US traffic, among 87 advertisers that bought both paid ChatGPT and paid Google search:

• Revenue per visit was $0.89 from paid ChatGPT, against $2.45 from paid Google search.
• Conversion rate was 0.37%, against 1.17%.
• 73 of the 87 advertisers (84%) recorded no paid ChatGPT sale at all. The other 14 accounted for all 570 transactions.

The authors estimate that, for ChatGPT clicks to match search on profit, they would need to cost around $0.26, far below the $1.72 paid by the largest publicly known account.

How much weight should you put on that?

Some, but not all. The authors list the limits themselves. Tracking captures only the last click, not the prompts or impressions before it. One advertiser accounts for 66% of the paid ChatGPT purchases in the financial sample. The paid results rest on about 450,000 sessions, mostly in the US, from the first 28 weeks of advertising. And a last-click view will undercount a channel that may work earlier in a purchase, the way a recommendation does.

OpenAI, for its part, cites individual advertisers with strong returns. Both can be true: a channel can work well for a few categories and poorly on average.

Why might conversational ads convert differently?

Two plausible reasons, which are our reading rather than findings of the study. First, a person asking ChatGPT for advice is often earlier in a decision than a person typing a product name into Google, so the intent is weaker at the moment of the click. Second, the answer itself may already satisfy the need, so the click that follows is exploratory. If either is right, ChatGPT ads should be judged more like upper-funnel media than like search.

How should you test ChatGPT ads?

• Ring-fence the budget. Treat it as a test with a fixed amount and end date, not as an extension of search.
• Measure beyond last click. Use a holdout (a market or audience that sees no ChatGPT ads) and compare total sales, branded search and direct traffic, not just attributed conversions.
• Pick categories where advice matters. Considered purchases, where people genuinely ask for recommendations, are the natural fit. Commodity products bought on price are not.
• Match the landing page to the conversation. A user arriving from a detailed exchange needs a page that continues it, with specifics and comparisons, not a homepage.
• Watch relevance. Check which conversations your ads appear in. Placement next to unrelated or sensitive topics is a brand risk as well as wasted spend.
• Set a kill threshold in advance. Decide the cost per incremental sale at which you stop, before the test starts.

The bottom line

Ads inside AI assistants are clearly where part of the market is going, and early-mover pricing can be an advantage. But the best public evidence so far says the average paid ChatGPT visit is worth roughly a third of a paid search visit. Test it like a new medium, measure it like one, and let the results, not the hype, set the budget.