
adMarketplace says advertisers in its AI chat ad program saw conversion rates rise 156 percent over the past year. The release behind that number gives no starting rate, no advertiser count and no click volume, so the 156 percent cannot be checked or compared to anything.
adMarketplace sells this inventory, and the data is its own. That does not make it wrong. It does mean the headline figure is a sales number until someone shows the math. The more useful data sits lower in the release: the median query ran 8 words on browsers, 5 on comparison engines and 3 on Buy Now Pay Later (BNPL) apps. For anyone managing PPC (pay-per-click) budgets, that describes three different channels sharing one label.
If a rep pitches you AI search ads on the 156 percent figure, ask for the absolute conversion rate by publisher, before and after. In paid search audits, the first thing I check on a blended number is what got added to the blend.
What adMarketplace Released
The New York company published its findings on November 18, drawing on what it calls a year of live campaigns. The product launched as an alpha in October 2025. The release calls it AI Discovery. The company’s own site calls it AI Discover.
The main claims:
- Conversion rates for beta advertisers rose 156 percent over the past year
- More than 30 percent of AI chat queries can carry an ad, by adMarketplace’s own classification
- Median query length was 8 words on browsers, 5 on comparison engines and 3 on BNPL shopping apps
- Electronics led on browsers, apparel on BNPL apps and furniture on comparison engines
- Of browser queries running seven words or longer, 65 percent were transactional
The release opens by saying a majority of consumers use AI to shop. It cites no source for that. Co-CEO Adam J. Epstein sums up the company’s view:
Conversational search is exploding and the conversations are becoming more commercial.
The 156 Percent Problem
A percentage increase needs two numbers. The release supplies neither.
No Baseline
A conversion rate that moves from 0.5 percent to 1.28 percent is up 156 percent. So is one that moves from 2 percent to 5.12 percent. Those are very different media buys. The release does not say which end of that range applies, how many advertisers are in the beta, or how many clicks sit under the figure.
The Mix Changed
This is the part I would press on. adMarketplace’s April beta announcement says the alpha ran on one publisher, Opera’s AI assistant. Trade coverage of that launch named three beta partners: Opera AI, the shopping engine Dupe and the BNPL platform Sezzle. A browser, a comparison engine and a BNPL app. Today’s release names none of them, but the three channels line up.
Now look at the query lengths again. Someone typing three words inside a BNPL app is already shopping, with financing one tap away. Add that traffic to a base of eight-word browser chats and the blended conversion rate climbs. No ad has to get better for that to happen.
I cannot tell from the release whether that explains the 156 percent. Neither can you.
One more line deserves a second read. The release says the share of browser queries with seven or more words hit 65 percent at its high point, and calls that more than double the average length of a traditional search query. That compares a share to a length. And a high point is a peak, not an average.
What the Query Data Is Good For
The channel split is the finding I would act on.
A three-word query in a shopping app behaves like keyword search. You know how to buy that. An eight-word question in a browser assistant carries a budget, a preference and a use case, and it rewards an ad matched to all three. adMarketplace says 65 percent of those longer browser queries were transactional. That is one company’s label on its own traffic, and still a reason to test.
The categories point the same direction. Electronics on browsers, apparel on BNPL apps, furniture on comparison engines. One campaign with one message across all three will average out to mediocre.
Keep the Scale in View
This inventory is not ChatGPT. It is a browser assistant, a shopping engine and a payments app.
The wider market is smaller than the press releases suggest. Semrush data I covered this month put AI tools at 1.29 percent of U.S. desktop activity, against more than 9 percent for traditional search. I have also reported Invoca’s finding that calls from ChatGPT qualified 49 percent of the time. AI referrals can convert. Rankability’s sign-up data and Erlin.ai’s study of AI recommendations pointed the same way. An ad placed inside the chat is a different product, and it has to prove itself separately. Companies have been pitching that product for a stretch now; I wrote about Koah’s “AdSense for GenAI” launch more than a year ago.
The Antitrust Pitch
Epstein ties the data to the remedies in U.S. v. Google. He says adMarketplace wants to compete for roughly $100 billion in search advertising syndication as those remedies open access. The figure is his, and the release gives no source for it.
The legal background is real. Judge Amit Mehta ruled in August 2024 that Google held an illegal search monopoly. His final judgment, issued December 5, 2025, requires Google to syndicate search results and search text ads to competitors for five years, according to Winston Taylor’s analysis of the judgment. Both Google and the Justice Department have appealed parts of the case. An opening exists. A $100 billion one is a forecast.
Epstein closes by saying the winners will be the companies that deliver performance and keep consumer trust. I agree with him. Performance is measurable. So measure it.
What To Do Next
- Ask any AI ad vendor for conversion rates by publisher, with the starting figure and the click count
- Run the test on your own conversion tracking, with a capped budget and a fixed end date
- Split campaigns by channel, because a three-word app query and an eight-word chat question need different ads
- Compare cost per acquisition against your non-brand search campaigns, where the buyer is also new to you
- Hold the 2027 budget line small until your own numbers say otherwise
I think ads in AI chat will earn a line in the paid search budget. This release does not tell you how big that line should be.
Every new ad channel arrives with a triple-digit growth figure. The ones worth buying will hand over the denominator when you ask.