研究称美国Google搜索广告访问收益高于ChatGPT广告

In U.S. traffic, a visit from a ChatGPT ad brought in $0.89 in revenue, against $2.45 for a visit from a paid Google search ad, according to a working paper by Maximilian Kaiser and Christian Schulze.

The paper, dated October 2026 and under review, follows site traffic and sales through OpenAI’s country-by-country ad launch. Its paid-ad figures come from 102 advertiser sites with full sales tracking.

At paid search’s rate of return, the authors value a ChatGPT ad click at about $0.26. The largest publicly known account paid $1.72 per click.

OpenAI’s ads FAQ says ChatGPT Ads “does not yet have performance benchmarks across advertisers, industries, or campaign types.”

Paid ChatGPT Visits Vs. Paid Search

ChatGPT ad visits converted at 0.37 percent in U.S. traffic versus 1.17 percent with paid search. When the authors compared the two channels within the same websites, they found the conversion and revenue gaps still existed even after accounting for type of device used and week.

Among the 87 advertisers who purchased both ChatGPT ads and paid search in U.S. traffic, 84 had lower earnings per ChatGPT ad visit than with paid search. A total of 73 had zero sales generated by their ChatGPT ads (across 18,726 visits between them) while the remaining 14 generated all 570 ChatGPT ad purchases.

One advertiser is responsible for 66 percent of all purchases made via paid ChatGPT across the 102 sites. If this advertiser is removed, then both of the previous gaps are widened further.

The Parity Price

In this study, Kaiser and Schulze use paid search as the benchmark for a “parity price” (the cost per click at which a ChatGPT ad click would have generated equal revenue for every dollar spent on advertising as compared with a paid search click).

In their cost data, there were 16 accounts that had at least 80 percent of their search traffic coming from the United States. Their median paid search click cost was $0.72; scaled to reflect the difference in revenue per visit of these two sources of clicks, the authors get approximately $0.26 per ChatGPT click. Because one advertiser accounted for 76 percent of all U.S. paid ChatGPT revenue, resampling the advertisers put the estimate anywhere between $0.05 and $0.57.

The $1.72-per-click figure comes from one account’s 15 days of spend, as reported by Opascope, which runs paid media accounts for clients. It’s more than six times the parity price. OpenAI’s ads basics page recommends starting CPC campaigns at a maximum bid of $3 to $5.

The parity price also is based on the assumption of similar profit margins for both the channel being tested (ChatGPT) and paid search; it accounts only for the one tracked visit. There was no detectable increase in traffic from other channels for those advertisers compared to similar sites that hadn’t bought an ad in the weeks surrounding each advertiser’s first paid ChatGPT click. However, this test can only rule out large increases in traffic.

Organic Referrals After Ads Arrived

Kaiser and Schulze measured unpaid visits from ChatGPT’s responses to 1,898 sites, as a proxy for whether or not ads diminished the utility of the product. Organic ChatGPT referrals over the first 13 weeks post-launch changed by an estimated –5.9%, relative to organic search referrals to the same websites and compared to the EU, where there was no advertising. This estimate has a range of –12.6% to +1.2%, including no change. Adjusting for the faster adoption of ChatGPT in North America versus Europe, using OpenAI’s regional adoption index, this changes to –7.2% to –8.4%, with ranges that exclude zero. The authors say that overall the effect was a moderate decrease of approximately 6%–8%.

Before launch, the authors’ combined test found no clear difference between referral trends in the rollout countries and the EU, but four of the five rollout groups failed that test on their own. The same combined test also flagged 49 of 100 simulated rollouts in which no ads ran.

Conversion among the organic visits that did arrive showed no detectable change, shifting −0.02 percentage points from the 2.21% pre-launch baseline.

How The Study Worked

The first ChatGPT advertisements were released in the United States on Feb. 9, 2026. ChatGPT ads reached Australia, Canada, New Zealand, United Kingdom, Japan, Korea, Brazil and Mexico by early August. ChatGPT ads did not run in the European Union until Aug. 24, two days after the 60-week data window expired.

Kaiser is at the Vienna University of Economics and Business, and Schulze is at the Frankfurt School of Finance & Management. The data comes from Google Analytics e-commerce tracking provided free by Grips Intelligence, an e-commerce data company. Kaiser is a scientific advisor to Grips and previously worked there. The paper says the company had no role in the research and has no commercial relationship with OpenAI.

The data captures the last click before each visit, but it doesn’t include prompts, ad impressions, or users’ subscription tiers. For the paid side, advertisers chose to buy ChatGPT ads, so this comparison reflects their results rather than proving the ads caused them.

Why This Matters

You can calculate the parity price for your own account by dividing the revenue per visit from ChatGPT ads by the revenue per visit from paid search, then multiplying the result by your search cost per click. The result is the price at which ChatGPT clicks would match search on tracked revenue per dollar. It leaves out profit margins and any later purchases the last click doesn’t capture.

Before you compare, make sure to check your ChatGPT ad sessions for ad review crawler visits. According to OpenAI’s basics page, it is possible for crawlers of this type (ad-review) to visit landing pages using tracking parameters, and third-party analytics tools can flag these types of visits as paid. Other reported results are more favorable, though they were based upon different metrics than those used by the authors of the paper. On Aug. 31, OpenAI posted that an ecommerce advertiser experienced a return on ad spend of three times their investment over a period of 28 days. Additionally, according to DoubleVerify, the attributed cost per acquisition for Weight Watchers was 15.3% less than their benchmark.

Looking Ahead

The paper’s data ends Aug. 22. OpenAI has since expanded ChatGPT Ads to 89 countries and says more than 100,000 advertisers have run campaigns. Its Aug. 31 post said CPC and outcome-optimized bidding accounted for most campaigns.

The authors are suggesting that since ad clicks for ChatGPT appear as sponsored cards and therefore appear separately from the answer, they could be worth little. On top of this, if OpenAI were to move the ads in with the answer (i.e. remove the separation) the value of an ad click could go up but at the same time so could the number of people who distrust ChatGPT’s answers. OpenAI’s Oct. 21 post says its ads are labeled and kept apart from answers. The same post says the company is “continuing to explore new ways for businesses to interact with consumers in more native ways in ChatGPT.”

Featured Image: Tetiana Yurchenko/Shutterstock