Paid ChatGPT visits earned $0.89, Google ad visits $2.45: what a 1,898-shop study means for brands

Most of what brands have heard about ChatGPT ads so far has come from two places: OpenAI, which sells them, and a handful of advertisers and agencies willing to share one account's results. A new working paper from Maximilian Kaiser (Vienna University of Economics and Business and Grips Intelligence) and Christian Schulze (Frankfurt School of Finance & Management) is the first large outside look. It uses analytics data from 1,898 online shops that together sell about $16 billion a year, and it covers 60 weeks, from June 29, 2025 to August 22, 2026, or nearly 10 billion visitor sessions.

Search Engine Land covered it on November 2 under the headline number: paid ChatGPT visits generated 64% less revenue per visit than paid Google Search. We read the full paper and its appendix tables because the details change how a packaged goods brand should read that headline. Below are the figures we think matter, charted from the paper's own tables, with the caveats the authors attach to them.

How the study works

ChatGPT ads did not arrive everywhere at once. They launched in the United States on February 9, 2026, then in Australia, Canada and New Zealand on April 16, the United Kingdom on June 4, Japan and Korea on June 19, and Brazil and Mexico in early August. The European Union stayed without ads until after the data window closed. That staggered rollout gives researchers something close to a natural experiment: they can compare what happened to ChatGPT referrals in a market after ads arrived against the same weeks in Europe, where nothing changed.

The paper asks two questions. First, did ads make ChatGPT less useful to shoppers, which would show up as fewer organic referrals or referrals that buy less often? Second, are the ads worth what advertisers pay for them? The first question uses all 1,898 shops. The second uses a smaller paid panel of 387 sites that received paid ChatGPT traffic, of which 102 had complete revenue tracking.

One caveat applies to every paid number below. Advertisers chose to buy ChatGPT ads. Nobody randomized them. The authors say plainly that this part of the evidence is descriptive, not causal. It tells you how paid ChatGPT traffic performed for the stores that bought it, not what it would do for yours.

The headline: a third of the revenue per visit

In US traffic, where most ad delivery happened, the gap is wide on both of the numbers that make up revenue per visit.

Show the data

Channel	Value
Paid Google Search	2.45$
Paid ChatGPT	0.89$

Source: Kaiser and Schulze (2026), Table J.1 Panel B. Revenue as reported by each site's own analytics, including taxes and excluding shipping. Descriptive comparison: advertisers chose to buy ChatGPT ads.

The conversion gap explains most of it. Paid ChatGPT visitors bought 0.37% of the time. Paid search visitors bought 1.17% of the time. Within the same shops, ChatGPT orders were also smaller, by $74 on average after adjusting for site, device and week.

Show the data

Channel	Value
Paid Google Search	1.17%
Paid ChatGPT	0.37%

Source: Kaiser and Schulze (2026), Table J.1 Panel B. Pooled levels; the paper's adjusted within-site difference is -1.87 log-odds (p = .005).

The pattern is not one bad account dragging the average down. Of the 87 advertisers that bought both paid channels in US traffic, 84 (97%) earned less per ChatGPT visit and 85 (98%) converted ChatGPT visits at a lower rate. Most never made a sale from the ads at all: 73 of the 87 recorded no paid ChatGPT transaction across 18,726 paid ChatGPT sessions, and the other 14 sites carried all 570 transactions.

Measure	Paid ChatGPT	Paid Google Search
Revenue per visit	$0.89	$2.45
Conversion rate	0.37%	1.17%
Advertisers earning less per ChatGPT visit	84 of 87	n/a
Advertisers with no paid ChatGPT sale	73 of 87	n/a
Price per click	$1.72 (largest public account)	$0.72 (advertiser median)
Break-even ChatGPT click price at search returns	about $0.26	n/a

The paper also compares both channels against direct traffic, meaning people who typed the address or used a bookmark. Paid Google Search did not differ detectably from direct traffic on conversion, order size or revenue per visit. Paid ChatGPT was the only channel that fell short of direct traffic on all six measures the authors tracked, including bounce rate, pages per visit and time on site. Its visitors stayed 104 seconds less than direct visitors.

The price question

Revenue per visit only matters next to what the visit costs. The authors work out a parity price: the cost per ChatGPT click at which it would return as much per dollar as a search click. The advertisers in their cost sample paid a median of $0.72 per paid search click. Scale that by the revenue ratio (0.89 divided by 2.45, or 0.36) and a ChatGPT click is worth about $0.26 at the same rate of return. Resampling the advertisers gives a range of $0.05 to $0.57.

The largest publicly known ChatGPT account reported paying $1.72 per click, more than six times that parity price. The paper notes this sits toward the low end of what practitioners report, and below OpenAI's own suggested starting bid of $3 to $5 per click.

Two cautions keep this honest. The parity price assumes similar margins in both channels, and it only counts the tracked visit. If seeing a ChatGPT ad leads people to buy later through another route, the ad is worth more than the tracked click suggests. The authors looked for that: around each advertiser's first paid week they found no detectable lift in organic ChatGPT, direct, organic search or paid search traffic. That check is observational, so it cannot rule out a delayed effect, but it does not support one either.

The part brands should not skip: organic held up

The more useful finding for most CPG brands is about the free answers, not the ads. When ads arrived in a market, organic ChatGPT referrals fell moderately. The authors estimate a decline of about 6 to 8% relative to the untreated EU, after adjusting for faster ChatGPT growth in North America, and they say neither test establishes the decline with full confidence.

Conversion did not move. Organic ChatGPT visitors converted at 2.21% before the launches, and the estimated change after was minus 0.02 percentage points, with a standard error of 0.05. In plain terms, people who still clicked through from an organic ChatGPT answer bought as often as they did before ads existed. Over the full window, organic ChatGPT traffic converted at 1.99%, and before the rollout its conversion rate roughly matched organic Google search.

Show the data

Channel	Value
Organic ChatGPT, before ads launched	2.21%
Organic ChatGPT, full 60-week window	1.99%
Paid ChatGPT, US	0.37%

Source: Kaiser and Schulze (2026), Sections 2 to 4. Organic figures come from the 1,898-site panel; the paid figure from the 102 advertisers with revenue tracking. Different samples, shown side by side for scale.

Organic ChatGPT is still a small channel in this data. It made up 0.23% of all visits to the panel over the full window and 0.40% over the final four weeks. Small, but growing, and converting like search. Paid ChatGPT made up 0.030% of visits on the paid panel in the final four weeks.

What this means for a CPG brand

Most of the 1,898 shops are general online retailers, not grocery or household brands, and the paid sample is 102 sites. So treat this as the best outside evidence available, not a verdict on your category. With that said, three things follow.

The organic answer is still the asset. If shoppers who click through from an unpaid ChatGPT answer convert at search-like rates, the work that gets your product into that answer keeps paying. That work is unglamorous: product facts that match across your site, retailer listings and feeds; clear size, count and ingredient details; reviews; and third-party coverage the assistant can cite. Our guide on how ChatGPT ranks merchants and tip on checking whether ChatGPT recommends you are good starting points.

If you test paid, test small and against your own baseline. The paper's method is one any brand can copy. Tag paid ChatGPT clicks separately, then compare revenue per visit with your own paid search over the same weeks. Divide your ChatGPT revenue per visit by your search revenue per visit, multiply by what you pay per search click, and you have your own parity price. If the going ChatGPT price is several times that, you know what you are buying. Our GA4 setup guide for AI traffic covers the tagging side.

Low-priced, frequent purchases are the hardest case. A store that sells $200 items can survive a low conversion rate on a few large orders. A $6 snack cannot. The paper does not break results out by category, so we can only say the arithmetic is tougher for low-ticket goods, not that the data proves it.

What to watch next

The paper is a first version, posted in October 2026, and the authors say the companion page will carry updates. Three things could change the picture. Ad formats are still changing: OpenAI added conversion-optimized campaigns in June 2026, and on October 30 Search Engine Roundtable reported a wider standard ad card and multi-image ads in open beta, with ads now in 89 countries. Ads reached the EU on August 24, after the data window, so a second version could include a much bigger market. And the paid sample will grow as self-serve buying spreads.

We will keep checking how organic ChatGPT answers treat the products we follow, the way tools like SearchPanel from XENA Intelligence track shopper-style prompts at scale. For now the simplest reading is the authors' own: the free recommendations kept their value after ads arrived, and the paid clicks have a long way to go to match what search delivers.