ChatGPT’s pitch to advertisers is simple: put brands in front of a fast-growing, high-intent audience. A new working paper complicates that story. Paid visits from ChatGPT ads are generating a fraction of the revenue that paid Google Search visits produce, and most early advertisers are not seeing a purchase at all.
What the study found
Researchers from the Frankfurt School of Finance & Management and the Vienna University of Economics and Business analyzed 1,898 ecommerce sites with $16 billion in annual sales and nearly 10 billion visits over 60 weeks, from June 2025 through August 2026. Paid ChatGPT sessions generated an average of 89 cents in revenue, while paid Google Search sessions generated $2.45. Conversion rates were 0.37% for ChatGPT ads and 1.17% for Google Search.
- 82% of the 102 advertisers in the study recorded no purchases from their ChatGPT ads.
- Google Search delivered more than six times as much revenue per ad dollar at publicly reported prices.
- A ChatGPT click would need to cost about 26 cents to match search return; the largest known account paid $1.72 per click.
The paper remains under review, so the findings are early evidence rather than a settled verdict. Still, they challenge the assumption that AI search ads are a direct substitute for search ads.
Scale is racing ahead of proof
OpenAI says ChatGPT Ads crossed a $1 billion annualized revenue run rate less than 200 days after launch, and projects advertising revenue of $102 billion by 2030. The researchers estimate current ads bring in roughly $1 per free user a year while serving that free user costs about $5. Because OpenAI is privately held, the run-rate figure cannot be checked against public filings.
Advertiser adoption is easier to see: Adthena tracked 1,082 brands advertising in June across the U.S., U.K., Australia, and EU, rising to 3,741 in July and 6,333 in August. New Street Research estimated about 60,000 advertisers by early October, up from roughly 20,000 at the end of August. More accounts, however, do not confirm stronger returns. OpenAI’s claim that ChatGPT handles about 20% as many search-like queries as Google is also hard to verify without disclosed methodology; independent estimates put the figure closer to 7.5%.
Paid and organic AI referrals are different channels
The paper found a sharp split inside ChatGPT itself. Unpaid ChatGPT referrals converted at 2.21% before the advertising rollout, far above the 0.37% conversion for paid placements. ChatGPT also sent 6% to 8% fewer unpaid visitors to ecommerce sites after ads launched, with the larger estimate statistically significant after adjusting for faster North American usage growth. The study does not establish why referrals declined.
That does not make AI traffic worthless. Adobe Analytics found AI-referred traffic to U.S. retail sites generated 83% more revenue per visit than non-AI traffic in September, with 90% better conversion, 73% more time on site, and 25% more pages viewed. The datasets measure different things, but both point to one planning rule: paid AI ads and organic AI referrals should be tracked separately, not lumped together as “AI traffic.”
What marketers should do now
Treat ChatGPT ads as an experiment with its own break-even benchmark. If a click needs to cost roughly 26 cents to match search return, evaluate campaigns on revenue per session and conversion, not reach or impressions alone. Run controlled tests in one market before shifting budgets from proven search channels.
Attribution is still unsettled. Farhad Divecha, group CEO at Accuracast, told Digiday: “We continue to see discrepancies there.” Some issues are new-system teething problems, he said, while others come from messy attribution. Until those pipes clear, small tests and careful analytics beat sweeping budget moves.
Source: MarTech
