Paid ChatGPT traffic generates substantially less revenue per visit than paid Google Search traffic, according to new research.

Paid ChatGPT visits generated an average of 89 cents in revenue per session, compared with $2.45 from paid Google Search, according to researchers at the Frankfurt School of Finance & Management and the Vienna University of Economics and Business. Conversion rates were 0.37% for ChatGPT ads and 1.17% for Google Search.

Among the 102 advertisers in the study, 82% recorded no purchases from their ChatGPT ads. Researchers concluded Google Search delivered more than six times as much revenue per advertising dollar at the publicly reported prices.

The paper 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. ChatGPT introduced ads in the U.S. and then eight additional countries, while the EU remained ad-free and served as the control group. The paper remains under review, so the findings are early evidence rather than a settled verdict.

The researchers estimated that a ChatGPT click would need to cost about 26 cents to match paid search’s return. The largest publicly known account in their data paid $1.72 per click.

The problem extends beyond advertiser returns. OpenAI’s own projections assume advertising revenue reaches $102 billion by 2030, or about $40 per free user. 

The company says ChatGPT Ads reached a $1 billion annualized revenue run rate less than 200 days after launch. OpenAI has not provided evidence supporting that figure, and because it is privately held, the claim cannot be checked against public financial filings. The number is also an annualized run rate, not $1 billion in ad revenue already collected.

The researchers estimate current ads generate about $1 per free user annually while serving a free user costs roughly $5. More advertisers and more clicks won’t close that gap unless each click becomes much more valuable.

OpenAI’s scale claims difficult to verify

OpenAI claims ChatGPT handles roughly 20% as many “search-like” queries as Google.

Applied to Google’s last disclosed annual search volume, that would imply more than 1 trillion ChatGPT search-like queries a year. But OpenAI hasn’t disclosed the underlying count or explained enough about how those queries are measured to establish a clean comparison with Google Search.

Another estimate comes in much lower. SEOClarity calculated ChatGPT’s search volume at about 7.5% of Google’s, which would put it at about 375 billion search-like queries annually.

10X your SEO with Semrush for Enterprise.

The world’s most powerful SEO platform, purpose-built for Enterprise.

Request demo

Advertiser growth is easier to observe. Adthena tracked 1,082 brands advertising across the U.S., U.K., Australia, and EU in June, rising to 3,741 in July and 6,333 in August. New Street Research estimated ChatGPT had roughly 60,000 advertisers by early October, up from about 20,000 at the end of August.

Those figures show adoption. They don’t show campaign returns.

Paid traffic looks very different from organic AI referrals

The working paper found a sharp split inside ChatGPT itself. Paid visits converted at 0.37%, while unpaid ChatGPT referrals converted at 2.21% before the advertising rollout. Visitors arriving through paid ChatGPT placements also performed worse than those arriving through Google ads, Google search, direct traffic, and ChatGPT’s own unpaid recommendations.

Other research on unpaid AI traffic points in the same direction.

Adobe Analytics found AI-referred traffic to U.S. retail sites generated 83% more revenue per visit than non-AI traffic in September. Those visitors converted 90% better, spent 73% more time on retail sites, and viewed 25% more pages.

The datasets measure different things. Adobe looked at AI referrals broadly, whereas academic research focused on paid ChatGPT visits. But both reinforce the need to separate paid AI traffic from organic AI referrals rather than treating “AI traffic” as one channel.

Organic referrals fell after ads appeared

Paid performance wasn’t the only change the researchers found.

ChatGPT sent 6% to 8% fewer visitors to the ecommerce sites after the ads launched, while people who still arrived via unpaid referrals converted at roughly the same rate as before.

The initial estimate put the decline at 5.9% during the first 13 weeks after ads appeared, compared with markets without ads, although that result wasn’t statistically significant. After adjusting for faster growth in ChatGPT usage in North America than in Europe, the estimated decline rose to between 7.2% and 8.4% and became statistically significant under those assumptions.

The study does not establish why referrals declined. But the finding matters because ChatGPT’s current ad design already keeps sponsored placements outside the answer. Free and Go users see a labeled “Sponsored” card below ChatGPT’s buying advice, rather than an ad inserted into the recommendation itself.

The researchers argue that OpenAI faces a difficult tradeoff. Better targeting or placing ads more directly inside answers could make the inventory more valuable, but could also put more pressure on the product experience that generates those unpaid referrals.

Advertiser adoption is moving faster than advertiser proof

ChatGPT’s advertiser base is spreading beyond retail. Sensor Tower data cited by Digiday shows shopping’s share of U.S. ChatGPT ad spending fell from 38% in April to 20% in September, while software rose from 13% to 18%, financial services from 2% to 13%, and travel and tourism from 5% to 9%.

Measurement is still unsettled.

“We continue to see discrepancies there,” Farhad Divecha, group CEO at Accuracast, told Digiday. “Some are teething problems because it’s a new system that has to be set up on advertiser systems, but some problems are down to attribution being messed up.”

OpenAI can point to a growing advertiser base and enormous usage. The independent performance data available so far is much less impressive: lower conversion rates, lower revenue per visit, weaker return per advertising dollar, and unresolved attribution problems.

The new research adds another problem for OpenAI. Its current ads appear too weak to support the advertising revenue the company projects, and making them more valuable could require product changes that risk the unpaid traffic advertisers already receive.