ChatGPT ads return 64% less revenue per visit than Google search, new research finds

A new working paper puts a number on ChatGPT advertising's return, and it is weak: across 102 U.S. advertisers with full sales tracking, a paid ChatGPT ad visit generated $0.89 in revenue, against $2.45 for a paid Google search ad visit — a 64% shortfall that holds up even after accounting for device type and week.

What the study measured

The paper, "Can Advertising Fund Free AI? Evidence from ChatGPT's International Advertising Rollout," comes from Maximilian Kaiser at the Vienna University of Economics and Business and Christian Schulze at the Frankfurt School of Finance & Management. It is dated October 2026 and currently under review. The authors tracked 1,898 online shops worth a combined $16 billion in annual revenue over 60 weeks, from June 2025 to August 2026, using data donated by Grips Intelligence, an e-commerce analytics firm. Kaiser serves as a scientific advisor to Grips; the paper states the company had no role in the research questions, design, analysis or writing, and has no commercial relationship with OpenAI.

ChatGPT ads launched in the U.S. on Feb. 9, 2026, and reached Australia, Canada, New Zealand, the U.K., Japan, Korea, Brazil and Mexico by early August. The European Union stayed untreated until Aug. 24, just two days before the data window closed, which let the authors use the EU as a control group against the countries that got ads.

Among the 87 advertisers in the sample who bought both ChatGPT ads and paid Google search ads in the U.S., 84 earned less revenue per ChatGPT visit than per search visit. ChatGPT ad visits converted at 0.37%, against 1.17% for paid search. Seventy-three of those advertisers recorded zero sales from their ChatGPT ads across 18,726 combined visits; the other 14 generated all 570 ChatGPT-ad purchases in the sample, and a single advertiser accounted for 66% of all of them.

What a click is actually worth

Using paid search's rate of return as a benchmark, the authors calculate a "parity price" — the cost per click at which a ChatGPT ad click would deliver the same return as a search click. That price comes out to roughly $0.26. The largest publicly known account, reported by the ad-management firm Opascope via Search Engine Journal, paid $1.72 per click — more than six times parity. OpenAI's own ads guidance recommends advertisers start cost-per-click campaigns with a maximum bid between $3 and $5, and its ads FAQ states the format "does not yet have performance benchmarks across advertisers, industries, or campaign types."

The study also found a cost to running ads at all. After ads launched in a given market, unpaid referral visits from ChatGPT fell by an estimated 6% to 8%, adjusted for ChatGPT's faster adoption in North America than Europe. Conversion among the organic visits that still arrived did not change in any detectable way — the drop was in volume, not in the quality of who showed up.

Who this affects, and what to do about it

The sample is e-commerce: online shops with trackable carts and same-session sales, not B2B software companies running longer consideration cycles. None of that changes the core finding in software's favor. A SaaS trial or demo request rarely converts in the same browsing session an ad click arrives in, which means the last-click revenue tracking the study relies on would likely understate ChatGPT ad performance for software even further than it does for retail — or simply fail to detect it on the timelines marketing teams usually measure.

That matters because plenty of software marketing teams have treated ChatGPT ads as a channel worth a pilot budget on the strength of OpenAI's user numbers alone, without first checking whether the traffic converts at a price that makes sense.

What's still unproven

The authors are careful about what their data does not show. Advertisers chose to buy ChatGPT ads, so the comparison reflects the results of whoever opted in, not proof that the ads caused the revenue difference. The data captures only the last click before a visit — it does not include ad impressions, prompts, or whether a user was on a paid ChatGPT subscription tier. And the concentration of purchases in a handful of accounts means the true average is noisy: resampling the advertiser pool put the parity-price estimate anywhere between $0.05 and $0.57.

What is harder to dismiss is the direction of every figure in the study: lower conversion, lower revenue per visit, and a price advertisers are currently paying that is several multiples above what the channel is shown to return. For a market still being sold partly on the promise that AI search can carry real advertising revenue, that is not a small gap to close.