A new working paper puts the first hard numbers on whether ChatGPT advertising can pay its own way, and the early answer is no. Paid visits from ChatGPT ads in United States traffic generated $0.89 in revenue per visit, against $2.45 for a visit from a paid Google search ad. The authors calculate that a ChatGPT click would need to cost about $0.26 to match paid search on return, far below the $1.72 paid by the largest publicly known account.
What the study measured
Maximilian Kaiser (Vienna University of Economics and Business) and Christian Schulze (Frankfurt School of Finance and Management) studied ChatGPT’s staggered advertising rollout, using the European Union, where ads had not yet run, as comparison. Their data cover 1,898 shops, nearly 10 billion sessions, and 387 advertiser websites, with revenue figures from 102 advertisers with full sales tracking.
Paid ChatGPT visits converted at 0.37 percent, against 1.17 percent for paid search. Of 87 advertisers who bought both channels, 84 earned less per ChatGPT visit than per search visit. Seventy-three recorded no sales from ChatGPT ads, across 18,726 visits, while the remaining 14 generated all 570 ChatGPT ad purchases in the sample.
The parity price, and why it is fragile
The $0.26 parity price is the authors’ benchmark, not a market price. They took 16 accounts with at least 80 percent of their search traffic from the United States, found a median paid search click cost of $0.72, and scaled it by the revenue gap between the two channels.
That benchmark is fragile because the sales are heavily concentrated. One advertiser accounted for 66 percent of all purchases from paid ChatGPT ads across the 102 sites. Resampling the advertisers to allow for that concentration puts the parity estimate anywhere between $0.05 and $0.57, a range the authors report openly.
The $1.72 figure is also a single-account observation, not an average price. Search Engine Journal reports it from one account’s 15 days of spend, as disclosed by the paid-media firm Opascope. OpenAI’s own guidance recommends starting cost-per-click campaigns at a maximum bid of $3 to $5, well above the parity range.
Did ads hurt the free product?
The paper also tests whether introducing ads devalued ChatGPT itself, using unpaid referrals as a proxy for product value. Organic ChatGPT referrals fell by an estimated 5.9 percent over the first 13 weeks after ads launched, relative to organic search referrals and to the European comparison group.
That headline estimate carries a wide range, from a 12.6 percent fall to a 1.2 percent rise, so it includes no change at all. Adjusting for faster ChatGPT adoption in North America than in Europe moves the estimate to a fall of 7.2 to 8.4 percent, with ranges that exclude zero.
Conversion among the organic visits that still arrived showed no detectable change, shifting by just 0.02 percentage points from a 2.21 percent pre-launch baseline. The authors therefore describe a moderate fall in volume, not in the quality of the visits that remained.
What the study cannot prove
The study is observational, and its limits matter as much as its headline. Advertisers chose to buy ChatGPT ads, so the comparison reflects the results of early buyers rather than proving the ads caused them. The data capture only the last click before each visit, missing later purchases, and they include no prompts, ad impressions, or subscription tiers.
The data also end on August 22, 2026, before ChatGPT ads reached the European Union on August 24 and before OpenAI’s later expansion. The paper is version one of an October 2026 working paper, under review rather than peer-reviewed, and its data were provided free by Grips Intelligence, a company for which Kaiser serves as scientific adviser.
Why It Matters
For advertisers, the practical question is not whether ChatGPT ads work in the abstract, but what price makes them worth buying. On these figures, paying search-like prices for ChatGPT clicks means paying several times the value the clicks returned in tracked revenue.
For OpenAI, the stakes are larger than one campaign. The paper’s central conclusion is that current ad performance is not sufficient to finance free access, which is the business model ads were meant to support. That finding lands alongside Google’s own admission that AI Mode ads earn less than search ads, suggesting the problem is not unique to one company.
Our read
Read carefully, this is a warning about price, not a verdict that AI advertising is doomed. Early buyers, early formats, and a single dominant advertiser all pull the averages in ways a mature market may not. OpenAI has since expanded ChatGPT ad formats, and its August post cited an ecommerce advertiser earning three times its investment over 28 days.
The more durable point is structural. If ads sit apart from answers, as OpenAI says they do, they may earn less attention than a link inside an answer, yet moving them closer risks the trust that makes people use the assistant at all. How OpenAI resolves that tension will decide whether the parity price rises or the product suffers.
Background
ChatGPT ads first launched in the United States on February 9, 2026, and reached Australia, Canada, New Zealand, the United Kingdom, Japan, Korea, Brazil, and Mexico by early August. OpenAI’s ads help pages say ChatGPT Ads does not yet have performance benchmarks across advertisers, industries, or campaign types.
Source: Kaiser and Schulze working paper and Search Engine Journal report
0 Comments