What did the new ChatGPT ads study find?
A paid click from a ChatGPT ad earned online shops $0.89 in revenue in the US, while a paid Google Search click earned $2.45. That is 64% less money per visit. The numbers come from a working paper by Maximilian Kaiser (WU Vienna) and Christian Schulze (Frankfurt School of Finance & Management), covered on November 2 by Search Engine Journal and PPC Land.
The researchers tracked 1,898 online shops that sell about $16 billion a year, over 60 weeks from June 2025 to August 2026, using Google Analytics ecommerce data. ChatGPT turned on ads country by country starting February 9, 2026, while the EU had none during the study window. That gave them a control group. The full paper and fact sheet are on the study's project page.
How bad were ChatGPT ad clicks compared to Google?
ChatGPT ad visitors in the US bought at a rate of 0.37%. Paid Google Search visitors bought at 1.17%, roughly three times as often. Of the 87 advertisers buying both channels, 84 earned less per ChatGPT visit, and 73 got zero sales from their ChatGPT ads across 18,726 visits between them.
| US traffic | Paid ChatGPT ad | Paid Google Search |
|---|---|---|
| Revenue per visit | $0.89 | $2.45 |
| Conversion rate | 0.37% | 1.17% |
| What a click is worth at Google's return | About $0.26 | $0.72 median paid |
The researchers then did the math advertisers care about. At the same return Google Search delivers, a ChatGPT click is worth about $0.26. The biggest publicly known ChatGPT account paid $1.72 per click, and OpenAI has suggested starting bids of $3 to $5, according to PPC Land. The paper's own summary: ChatGPT ads offer "more than six times worse value for money than Google Search ads."
Which ChatGPT traffic did convert well?
Free, organic recommendations from ChatGPT did. In US traffic, a visit from an unpaid ChatGPT answer was worth $3.27, the highest of the five channels the study measured, beating paid search. Across all countries, organic ChatGPT visitors converted at 1.31% versus 0.31% for paid ChatGPT ad visitors.
The authors' explanation is placement. ChatGPT's ads sit in a labeled sponsored card below the answer, apart from the advice people came for. When ChatGPT names your business inside the answer, it reads like a referral. Paid ChatGPT visitors viewed 1.4 fewer pages and left 104 seconds sooner than people who typed the site in directly.
Should a small business stop buying ChatGPT ads?
Not automatically, but you should stop judging them on clicks and cost per click. Judge every new ad channel on revenue per visit, or for a service business, booked calls and closed deals per 100 clicks. If ChatGPT can't hit your number after a fair test, move the money back to what works.
Keep the limits of this study in mind. It's a working paper still under review. The shops chose whether to buy ChatGPT ads, so it shows a gap, not proof the placement caused it. It uses last-click data, so a ChatGPT click that led to a purchase a week later through Google would count for Google. And it covers online stores, not plumbers, med spas or law firms. One advertiser also made up 76% of US paid ChatGPT revenue, which is why the $0.26 estimate has a wide range ($0.05 to $0.57). Kaiser is also a scientific advisor to Grips Intelligence, the company that supplied the data, which the paper discloses.
I've been running ads since 2013, and every new platform goes through this. Early clicks look cheap or exciting, then the money doesn't show up in the bank. When I test a new channel, I run it like this:
- Give ChatGPT ads a fixed test budget you can afford to lose, and set a max cost per click based on your own numbers.
- Tag every link with UTM parameters so ChatGPT leads land in your CRM under their own source.
- Ask every new lead where they found you, and log the answer. Last-click reports miss a lot.
- After three to four weeks, compare revenue per visit and cost per booked appointment against Google and Meta.
- Keep the channel only if it pays back on your numbers, not on the platform's dashboard.
To run that comparison fast, plug each channel's spend and revenue into our free ROAS calculator. A channel with cheap clicks and no sales will show up as a loss.
How do you get the ChatGPT traffic that actually buys?
You earn it the way you earn any referral: a strong review profile, a clear website that says what you do and where, and fast answers when people reach out. AI assistants pull from reviews, listings and your own pages when they recommend a business. Ads don't buy you a spot inside the answer.
Then the lead has to get a fast reply. A recommended business that takes four hours to call back loses the lead to the next name on the list, which we break down in our guide on why leads go cold.
Make every paid click count
Whatever channel a lead comes from, KenjiAI's Voice AI agents call or text them within a minute, ask where they found you, log the source in the CRM and book the appointment. You see revenue per channel instead of guessing from clicks. We've run campaigns at 5x ROAS, and fast follow-up on every lead is a big part of it.
Next step: pull last month's leads by source, work out revenue per visit for each channel, and decide where your next $1,000 of ad spend goes based on that number.
FAQ
Do ChatGPT ads work for small businesses?
Early data says they underperform. A November 2026 working paper found paid ChatGPT ad visits earned online shops $0.89 each in the US versus $2.45 for paid Google Search visits. Test them with a small fixed budget and judge them on revenue, not clicks.
How much is a ChatGPT ad click worth?
The study estimates about $0.26 per click at the same return Google Search delivers, with a range of $0.05 to $0.57. The largest publicly known account paid $1.72 per click, so most advertisers are likely overpaying for now.
Is organic traffic from ChatGPT valuable?
Yes. In the same study, a visit from an unpaid ChatGPT recommendation was worth $3.27 in US traffic, the highest of the five channels measured. Strong reviews, accurate listings and a clear website help you show up in those answers.
Who ran the ChatGPT ads study?
Maximilian Kaiser of WU Vienna and Christian Schulze of the Frankfurt School of Finance and Management, using Google Analytics data from 1,898 online shops supplied by Grips Intelligence. It is a working paper still under peer review, and Kaiser advises Grips, which the paper discloses.
Originally published at kenjiai.com. Published by the Media Traffics | KenjiAI team.
Top comments (0)