A click from an AI conversation does not necessarily signal readiness to buy. In U.S. traffic, a visit from a ChatGPT ad generated an average of $0.89 in website revenue, compared with $2.45 from a paid Google search ad. Conversion rates were 0.37% and 1.17%, respectively. These figures describe an early phase of ChatGPT Ads rather than a universal forecast for any campaign: the sample is limited, and the platform continues to evolve.
The findings come from research on advertising in ChatGPT by Maximilian Kaiser and Christian Schulze. The October 2026 working paper is under review. Its paid-traffic financial results cover 102 advertiser websites with complete sales tracking.
Averages Conceal the Gap between Advertisers
Google's advantage persisted when the researchers compared channels within the same websites while accounting for device and week. The observed gap therefore cannot simply be explained by one channel receiving smartphone users and the other desktop users, or by differences in timing.
The distribution of sales is particularly revealing. Of 87 advertisers buying both ChatGPT Ads and paid search for U.S. traffic, 84 earned less revenue per ChatGPT ad visit. For 73 businesses, the channel generated no sales at all, despite a combined 18,726 visits. The remaining 14 websites accounted for all 570 purchases.
A few large participants also shaped the overall result. One advertiser generated 66% of paid ChatGPT purchases across the full financial sample of 102 websites. Removing that advertiser made ChatGPT's conversion and revenue-per-visit shortfalls larger.
OpenAI itself acknowledges in its ChatGPT Ads FAQ that it does not yet have performance benchmarks across advertisers, industries, or campaign types. The study's averages can help businesses check their expectations, but they cannot replace account-level evidence.
What the Study Could Measure
The study covers 60 weeks ending August 22, 2026. ChatGPT ads began appearing in the United States on February 9. By early August, they had also reached Australia, Canada, New Zealand, the United Kingdom, Japan, South Korea, Brazil, and Mexico. Ads started in the European Union on August 24, after the observation window closed, allowing EU countries to serve as a control group.
Kaiser is affiliated with the Vienna University of Economics and Business, and Schulze with Frankfurt School of Finance & Management. Grips Intelligence supplied Google Analytics e-commerce data free of charge. Kaiser previously worked for the company and now serves as its scientific adviser. The paper states that Grips had no role in the research and has no commercial relationship with OpenAI.
The analytics capture the last click before a visit. They do not reveal user prompts, ad impressions, or subscription tiers. Businesses also chose for themselves whether to buy ChatGPT Ads. The comparison therefore describes those advertisers' results; it does not, by itself, establish that the advertising platform caused the differences.
What Click Price the Revenue Supports
To compare channel economics, the researchers calculated a parity price: what a ChatGPT click could cost while generating the same tracked revenue per advertising dollar as paid search.
Within the cost-data sample, 16 accounts received at least 80% of their search traffic from the United States. Their median paid search CPC was $0.72. Adjusting for the ratio of revenue per visit between the channels produced a ChatGPT parity price of roughly $0.26.
That estimate carries considerable uncertainty. One advertiser generated 76% of U.S. paid ChatGPT revenue. Resampling advertisers produced estimates ranging from $0.05 to $0.57.
For comparison, Opascope's 15-day case study reported an actual CPC of $1.72 for one account, more than six times the $0.26 estimate. That is one account's result, not a platform-wide average. In its advertising basics guide, OpenAI recommends starting CPC campaigns with a maximum bid of 5. A maximum bid sets an auction limit; it is not the actual price of every click.
The parity calculation assumes comparable profit margins and counts only the tracked visit. Purchases made later through another channel are excluded. Around advertisers' first paid ChatGPT click, the researchers found no detectable traffic increase in other channels relative to similar websites that had not bought ChatGPT ads. However, the test was sensitive enough to rule out only large increases, leaving moderate delayed effects unresolved.
What Happened to Unpaid ChatGPT Referrals
The researchers separately tracked organic ChatGPT referrals to 1,898 websites to investigate whether advertising reduced the product's usefulness.
During the first 13 weeks after launch, unpaid referrals changed by an estimated −5.9%, relative to organic search referrals to the same websites and the EU control group. The interval of −12.6% to +1.2% includes no change, so the initial model does not provide clear confirmation of a decline.
After adjusting for faster ChatGPT adoption in North America than in Europe, the estimated decline became −7.2% or −8.4%, with intervals excluding zero. The broader description of a 6–8% decline therefore depends on assumptions about regional adoption; it is not a single precise measurement.
There are also questions about how well countries' traffic trends could be compared. A combined pre-launch test found no clear trend divergence, but four of the five rollout groups failed that test individually. The same combined test flagged 49 of 100 simulated rollouts in places where no ads had appeared. These results limit confidence in a causal explanation for the traffic decline.
Conversion among the organic visitors who did arrive showed no detectable change. It shifted by −0.02 percentage points from a pre-launch baseline of 2.21%. The data do not show that these visitors became less likely to buy after advertising launched.
How to Apply the Findings to Your Campaigns
An account-specific CPC reference can be calculated by dividing ChatGPT Ads revenue per visit by paid search revenue per visit, then multiplying by paid search CPC. The result is the price at which the channels would deliver equal tracked revenue per advertising dollar. Assessing profitability also requires margins and subsequent purchases.
Before calculating, check the quality of the sessions being counted. OpenAI warns that ad-review crawlers may visit landing pages with tracking parameters. Third-party analytics can classify those visits as paid traffic, although they are not billable advertising clicks.
Positive cases exist as well. In its August 31 update, OpenAI described an e-commerce advertiser achieving a ROAS of 3 over 28 days. DoubleVerify reported that Weight Watchers' attributed acquisition cost was 15.3% below its own established benchmark. Those results use different metrics and measurement approaches, so they cannot be compared directly with revenue from a single visit.
Rapid Platform Changes Limit How Far the Findings Apply
The findings apply to the period ending August 22. According to OpenAI's October 21 announcement, ChatGPT Ads had expanded to 89 countries, and more than 100,000 advertisers had run campaigns. The company had already said on August 31 that CPC and outcome-optimized bidding accounted for most campaigns.
The researchers suggest that placing sponsored cards below the answer may limit the value of ad clicks. Closer integration into the answer could theoretically increase that value, while also creating a risk to user trust. That remains a hypothesis: on October 21, OpenAI reaffirmed that ads are labeled and separate from answers, while saying it was exploring more native ways for businesses to interact with consumers within ChatGPT.





Комментарии