Who Pays for ChatGPT Ads? Tracing the $25B Budget Shift

The original headline on this page put a $25 billion figure on ChatGPT ads. It came from Evercore ISI analyst Mark Mahaney, who wrote in January 2026 that “a path to generating several billion dollars in ad revenue in 2026, going to $25B+ by 2030, seems reasonable,” as Business Insider reported. Three months later, OpenAI’s own investor projections, reported by Axios and Reuters, put ad revenue at $2.5 billion in 2026, $11 billion in 2027 and $25 billion in 2028, rising to $100 billion by 2030.

So the $25 billion arrives two years earlier in OpenAI’s plan than in the analyst’s. Either way, it is a forecast. What OpenAI had actually reached by August 2026 was a $1 billion annualized run rate. The useful question for anyone running a media budget is the one the forecasts skip: if advertisers spend $25 billion a year inside ChatGPT, whose budget does it come out of?

The $25 billion is several different numbers

The same round figure appears across AI advertising coverage, attached to different claims. Evercore’s is OpenAI ad revenue by 2030. OpenAI’s investor deck, as reported, reaches it in 2028. Earlier, The Information reported internal OpenAI figures showing nearly $25 billion from “free user monetization” by 2029, as MarTech summarized. And eMarketer forecast that US spending on AI-driven search ads across all platforms would reach nearly $26 billion by 2029, about 13.6% of all US search ad spending, which includes Google’s own AI formats.

These are not the same claim, and they should not be quoted as if they were. The first three are about one company’s revenue. The last is about a format across the whole market. Treat all of them as direction, not plan. The direction is clear enough: forecasters expect a meaningful share of search-style money to move into AI answers within a few years.

So far, the money comes from display and test budgets, not search

OpenAI has deliberately avoided asking for search budgets. In May 2026, an agency executive told Digiday that OpenAI’s advice was: “don’t take from search, but take it from anything else you’re doing up the funnel.” OpenAI’s Asad Awan said ChatGPT was “not the same as a discovery platform feed or pure search.” Another agency said clients preparing to spend were drawing from programmatic display budgets. Separately, Digiday reported that The Knot was funding its ChatGPT pilot from its test-and-learn budget.

The incumbents’ results agree. In the second quarter of 2026, Google Search and other revenue was $63.3 billion, up about 17% from a year earlier, and Meta’s advertising revenue was $59.4 billion, up 27%. A $1 billion annual run rate is roughly $250 million a quarter, about 0.4% of Google Search’s quarterly revenue on that basis. Whatever ChatGPT ads are taking, it is not yet visible in either company’s numbers.

That is not an accident. Money moves to a new channel through the budget line with the weakest proof requirements. Search budgets are held to cost per acquisition and return on ad spend, measured every week. A search manager who moves money to an unproven channel and misses target has to explain it. Display, prospecting and experimental budgets are judged more loosely, on reach, engagement or learning, so they can absorb a channel that cannot yet prove its results.

That was ChatGPT’s position for most of 2026. Conversion tracking only arrived in May, and Digiday noted that OpenAI was pushing cost-per-click ads while steering advertisers toward test budgets because it could not yet show what happened after the click. One agency executive put the consequence plainly: performance advertisers are “not going to be OK in a world where they’re not able to see post click activity.”

Experimental budgets are not small. One agency quoted by Digiday described splitting client spend 80/20 between core media and testing, with room to move toward 25%, and pet food brand Pawco dedicates 15% to 20% of its marketing budget to experiments. That is the pool ChatGPT is drawing from.

It also explains the forecasts. Test budgets can carry a channel to a few billion dollars. Getting to $25 billion requires the channel to win money that is held to performance targets, which means search and social. That move happens only if ChatGPT ads prove they bring customers at a cost comparable to the lines they replace.

Which budget lines to move, in what order

When a ChatGPT test works, it should graduate in an order that protects what already performs. Start by funding the test from the line it most resembles in function: prospecting and programmatic display, which reach people before they have decided and are usually the least well measured part of a plan.

Move non-brand search money only after a holdout test shows that ChatGPT brings new customers at a cost close to your non-brand search, for the same products. Research-heavy categories, where people increasingly ask an assistant before they search, are where that is most likely to happen first. Running that comparison properly is the work of performance marketing built on incrementality tests.

Leave brand search and retargeting alone. Brand search captures people already looking for you, and retargeting works on people who already know you. ChatGPT ads replace neither.

And keep a line for the part no budget buys. In an assistant, the most valuable position is inside the answer, and that is earned by being the source the answer draws on. That organic work, covered in how to get your brand cited in AI answers, belongs alongside search engine optimization for AI answers as well as Google. What the paid placement includes, and whether it is worth testing at all, is in the buyer’s guide to ChatGPT ads.

The $25 billion will come from somewhere if it comes at all. Advertisers decide where, one test at a time, and the ones who move money in the right order will not have to explain a missed target to anyone.

Frequently Asked Questions

Where does the $25 billion ChatGPT ads figure come from?

Several sources use it. Evercore ISI analyst Mark Mahaney estimated in January 2026 that OpenAI’s ad revenue could reach $25 billion or more by 2030. OpenAI’s own investor projections, reported by Axios in April 2026, put ad revenue at $25 billion in 2028 and $100 billion by 2030. Separately, eMarketer forecast that US AI search ad spending across all platforms would reach nearly $26 billion by 2029. All are forecasts, not results.

How big is ChatGPT’s ad business today?

OpenAI said in August 2026 that ChatGPT Ads had reached a $1 billion annualized revenue run rate in less than 200 days after launch, with tens of thousands of advertisers. That is about $250 million a quarter. For comparison, Google Search and other revenue was $63.3 billion in the second quarter of 2026 alone, so ChatGPT’s ad business is still a small fraction of the search market.

Will ChatGPT ads take budget from Google search?

Not yet. OpenAI has advised advertisers to fund tests from budgets other than search, and agencies report pilots funded from display and test-and-learn budgets. Google Search revenue grew about 17% year on year in the second quarter of 2026. For ChatGPT ads to reach the scale OpenAI projects, they will eventually need to win performance budgets, which depends on proving they bring new customers at a competitive cost.

Which budget should fund a ChatGPT ads test?

Fund it from prospecting, programmatic display or an experimental budget, not from brand search or retargeting. Those lines play a similar role and carry looser performance targets. Move non-brand search money only after a holdout test shows ChatGPT delivers new customers at a cost close to non-brand search for the same products. Measure with OpenAI’s pixel or Conversions API, or your own conversion data.