On 17 August 2026 Similarweb opened up competitive data on advertising inside AI answers, adding AI Ads to its Ad Intelligence line. The product covers ChatGPT, Google AI Mode and Google AI Overviews, and it is built on conversations from a real user panel rather than synthetic prompts fired at the models. The numbers that came with the launch describe a channel that is already crowded, converts poorly on first contact, and remains almost impossible to observe from the outside.
The density is already high
Twenty six percent of ChatGPT responses now carry a sponsored placement, shown to users on the Free and Go tiers. Nearly thirty percent of ad eligible queries in Google AI Mode display ads. More than forty percent of Google searches now trigger an AI Overview, and those frequently carry advertising too.
Anyone still describing AI advertising as experimental should sit with those three figures. A quarter of ChatGPT answers and nearly a third of eligible AI Mode queries is not a pilot. It is a live channel that most brands are simply absent from.
Two thirds arrive after the second prompt
The most instructive number in the release is easy to skim past. 66.3 percent of ChatGPT ads appear after the second prompt or later. The advertising is not attached to the opening question. It shows up once the conversation has developed.
That reshapes what a good ChatGPT campaign looks like. In paid search the whole transaction happens at the query: someone types, the auction runs, the ad appears. Here the exchange has to build first. The user states a need, the model responds, the user refines, and only then does the commercial moment arrive, by which point the model is holding a much richer picture of what the person actually wants. Writing creative for the opening question of a conversation means writing for a third of the inventory.
The click through rate is the counterweight. Similarweb puts ChatGPT ads at 0.50 percent, roughly a tenth of what a healthy paid search campaign returns. That is worth stating plainly rather than glossing over. The argument for the channel is not efficiency today; it is position in a channel whose economics are still forming.
No ad library, no reporting line
The transparency gap is the part with strategic consequences. None of these three environments publishes anything comparable to the Meta Ad Library or Google’s own Ads Transparency Center. There is no searchable record of what ran, who paid for it, or the context it appeared in.
It goes further than competitive blindness. Google Ads reporting does not separate AI Overview placements from other top of page positions. A media buyer running Search, Shopping or Performance Max can be serving inside a Google AI answer with no line in any report telling them so. The brand is present in the channel without knowing it is present in the channel.
Harel Amir, who runs Ad Intelligence at Similarweb, framed the opportunity directly: “This is the rare moment when a major ad channel is still wide open. The advertisers who can see what’s happening with AI ads now will have a real head start.”
For GEA planning the practical conclusion is narrow but firm. Because the platforms do not report placement or competitive context, the visibility has to come from somewhere else: third party panel data, and your own accumulated campaign history. Both take time to build, which is the strongest available argument for starting before the channel becomes legible to everyone at once.
Sources: PPC Land, Similarweb opens AI ad data as 26% of ChatGPT replies carry sponsored ads · Search Engine Land, Google AI Mode ads reach nearly 30% of queries