ChatGPT ads revenue per visit just got measured against Google Search ads, and the gap is bigger than most advertisers assumed. A new working paper tracking 1,898 online shops through OpenAI's ad rollout found $0.89 per ChatGPT ad visit against $2.45 for a Google Search ad visit — a 64% shortfall. If you're spending real money testing ChatGPT ads right now, this is the number to check before your next budget review.
What did the ChatGPT ads revenue-per-visit study find?#
Economists Maximilian Kaiser and Christian Schulze tracked sales data across 1,898 shops — about 10 billion sessions over 60 weeks — using free e-commerce tracking data from Grips Intelligence, as Search Engine Journal reported on 2 November. Of 87 advertisers running both channels in U.S. traffic, 84 earned less per ChatGPT visit than per search visit, and conversion ran 0.37% against 1.17% for search. One advertiser accounted for 66% of all ChatGPT purchases in the dataset — a detail that matters more than the headline number does.
We read the working paper itself on ecommerceresearch.org the morning it circulated. It's dated October 2026 and still under review, not peer-reviewed yet. Worth saying plainly.
Does this change anything if you're not running ChatGPT ads yet?#
If you haven't touched ChatGPT ads, nothing here changes your Tuesday. Keep reading anyway, because the "why" matters more than the number.
If you're running under roughly $2,000 a month in paid search or Meta spend, I wouldn't touch ChatGPT ads on this data. The parity price the paper estimates — what a ChatGPT click would need to cost to match Google Search economics — sits around $0.26. Some advertisers are reportedly paying north of $1.70 per click. That's not a rounding error, that's a different product.
If you're above that spend level and already diversifying — the way a lot of our GCC and European clients are after Meta's ad costs jumped 23% this quarter — a small, bounded test still makes sense. Just don't move budget off a channel that's working.
The honest trade-offs here#
In fairness to ChatGPT ads: this is an early-rollout study, U.S.-only for the revenue numbers, and self-selected advertisers skew the comparison — businesses that chose to test ChatGPT ads aren't a random sample. The paper's authors say as much. And OpenAI has cited a 3x ROAS case study of its own, so somebody's making it work.
Against it: the conversion gap held after adjusting for device and week, 73 of 87 dual-channel advertisers had zero paid ChatGPT sales across nearly 19,000 visits, and the study found a real, if modest, drop in organic ChatGPT referrals too — somewhere in the 6-8% range. I'll admit I expected the referral number to be worse than that. It wasn't.
How we're handling this in client ad accounts#
At WebEpex we manage Google Search, PMax and Meta budgets for service businesses across the GCC, Europe, the USA, Canada and India, and we built a one-line check into every account audit this week: is any budget sitting in ChatGPT ads, and could that same dollar do more in a channel we can already measure properly. For most accounts right now, the answer is move it back.
We're not writing ChatGPT ads off completely — we rebuilt our own attribution checklist around it the same week the PMax brand filter change landed, specifically so a client can run a capped test without it quietly cannibalizing search budget. That's the part agencies skip: test budgets need their own tracking, not a shared pool with the channel that's already converting.
Separately, if your organic traffic from ChatGPT matters to you, the referral drop ties into the same AEO work we cover in our piece on AI citations and earned media — getting cited well still beats paying for the click.
What I'd tell a client asking about this#
Three things, in order:
- Leave your core Google Search and Meta budget alone this week. Nothing in this study tells you to move it.
- If you want to test ChatGPT ads anyway, cap it at 5% of monthly spend and track it on its own conversion path, never blended with search. Give it 60 days, not two weeks.
- Under $2,000 a month in total paid spend, skip the test. The parity math doesn't work yet at that scale, and your time is worth more than the experiment.
If you want a second pair of eyes on where your ad budget's actually going versus where it's converting, send me your account and I'll tell you straight — two minutes, no pitch attached. cal.com/webepex/growth-review