Early data show ChatGPT Ads trail Google Search on revenue per visit
Early performance data point to a gap between ChatGPT Ads and Google Search advertising in revenue and transactions per visit, but they do not establish whether advertisers are shifting budget from one platform to the other. For marketers weighing an allocation change, the evidence also carries important limits: the comparison is descriptive, the channels’ click prices were not matched, and campaign attribution methods differ.
OpenAI launched ChatGPT Ads in February. The company said in an Oct. 21 engineering post that the product had passed a $1 billion annualized run rate and that more than 100,000 advertisers had run ads since launch. Those figures signal growth, but do not show how much spending came from Google Search or other channels.
Alphabet reported $63.35 billion in third-quarter revenue from Search and other activities, up 12%. On its earnings call, Chief Business Officer Philipp Schindler cited “the lower monetization we were seeing in AI Mode” as one factor in slower Search growth, without providing a figure or baseline. A Bank of America survey of 117 advertisers and agencies, as reported by TheStreet, identified Google Search as the most-mentioned potential source of future ChatGPT Ads spending. The survey captured plans, not completed budget transfers.
What the pooled performance data measures
A working paper by Kaiser and Schulze examined data from 102 websites with transaction and revenue tracking, provided by Grips Intelligence. For U.S. traffic, it reported $2.45 in revenue per session from paid search, compared with $0.89 from ChatGPT ad sessions. Transactions per session were 1.17% for paid search and 0.37% for ChatGPT Ads. The pooled figures cover 60 weeks through Aug. 22; the U.S. analysis included up to 28 weeks of ChatGPT ad sessions.
The researchers used a statistical model controlling for website, device type and week. They found lower conversion and revenue per session rates for ChatGPT ad sessions, and said the results held after removing individual advertisers. Among 87 advertisers that used both channels in U.S. traffic, 84 recorded less revenue per session from ChatGPT Ads.
The authors characterize the work as descriptive rather than causal. Advertisers selected whether to run ChatGPT ads, and the study does not test matched campaigns or prove that moving a specific advertiser’s budget would improve or weaken results. The researchers also note that about 20% of the websites were in services, software, education or finance, where conversions may include bookings, leads, licenses or sign-ups. The paper’s appendix describes a transaction as a completed order but does not explain how that definition relates to those other conversion types.
Grips Intelligence donated the data. The authors disclosed that Kaiser, a former Grips employee and current scientific advisor, had received salary and virtual company shares from the company; they said Grips played no role in conducting the research.
Why the value-per-dollar comparison is less settled
A Frankfurt School statement said advertisers received $0.52 in revenue for each dollar spent on ChatGPT Ads, compared with $3.40 for Google Search. Those figures were not directly measured campaign outcomes. They combine the study’s revenue-per-session results with a $1.72 ChatGPT cost per click reported by one agency account and a $0.72 median Google cost per click across 16 mostly U.S. accounts. The prices were not matched to the revenue generated by each website.
On that basis, a ChatGPT ad visit generated about one-third the revenue of a paid search visit while costing about 2.4 times as much. The $1.72 price came from Opascope, which reported roughly $60,000 in spend over 15 days on campaigns it scaled after early promise. The agency said the result was not representative of every test it launched and reported a 1.49x return on ad spend. OpenAI’s suggested starting maximum bids were cited as $3 to $5, but a maximum bid is a cap, not necessarily the price paid per click.
OpenAI told the paper’s author that the research reflected an earlier development period and extrapolated performance from a small sample. The company also pointed to product launches since the ads debuted, including Target CPA and Target CPC. Its FAQ says it does not currently provide performance benchmarks.
Agency tests underscore the need for comparable outcomes
Individual agency reports vary. StubGroup compared 30 days of campaigns for a B2B compliance company, reporting a $0.76 ChatGPT cost per click and a $92.21 cost per lead, versus $10.46 and $157.57, respectively, for Google Search. But Google’s conversion rate was 6.64%, compared with 0.8% for ChatGPT. The agency said the Google leads appeared more likely to become customers, while noting that a long sales cycle meant the eventual cost per customer was not yet known.
Workshop Digital reported a $3.14 ChatGPT cost per click, a 1.1% click-through rate and no qualified leads after spending $2,319. Its Google comparison showed a $6.76 cost per click, a 5.2% click-through rate and two qualified leads, though the agency called the campaigns “not a direct comparison.” Brafton reported spending $1,524.11 on a ChatGPT click campaign from June 17 to Aug. 28, generating 263 clicks and one form submission. Its Google and LinkedIn comparisons were external benchmarks.
OpenAI also highlighted a Jotform example reporting a cost per signup approximately 30% below paid search, based on the company’s internal July data. The example did not provide underlying costs or explain sign-up attribution; sign-ups are not equivalent to sales.
Set attribution rules before shifting budgets
OpenAI Ads Manager conversion totals may include eligible view-through conversions, and advertisers can choose a 7-, 14- or 30-day click window. The view-through window can be set to one day or disabled; both windows use last-touch logic. The study tracked ad-attributable visits, while Ads Manager reports clicks and analytics platforms report sessions. Redirects, consent settings, browser blocking and ad-review crawlers can contribute to discrepancies between those measures.
For a meaningful test, marketers should compare actual spend and attributed revenue across channels over the same period, document attribution settings, and report session or click volume alongside outcomes. Lead-generation teams also need consistent definitions of a qualified lead and enough time to assess conversion to customers. New ChatGPT campaigns have offered Target CPA for cost-target conversion campaigns and Target CPC for click campaigns since Nov. 9; Intelligent bidding was previously called Maximize Results. The research data predates these options and bid adjustments, so results generated under different strategies may not be directly comparable.
