Similarweb today released AI Ads, an Ad Intelligence dataset covering paid placements inside ChatGPT, Google AI Mode and Google AI Overviews, three surfaces that carry advertising without any public ad library behind them.

The announcement, issued from Tel Aviv over Business Wire at 8:00 a.m. Eastern Daylight Time on August 17, 2026, describes AI Ads as the newest addition to the company's Ad Intelligence line. According to Similarweb, customers can use it to run competitive analysis of advertising on ChatGPT, Google AI Mode and Google AI Overviews, worldwide and for specific markets.

Three figures anchor the release. According to Similarweb, 26% of ChatGPT responses already carry a sponsored ad, shown to users on the ChatGPT Free and Go tiers. Nearly 30% of ad-eligible Google AI Mode queries already show ads. And over 40% of Google searches now trigger an AI Overview, with those AI Overviews often including ads.

None of those three environments publishes a searchable record of what ran, who paid for it, or in what context it appeared.

The transparency layer that does not exist

The comparison Similarweb draws is with the tooling that grew up around the previous generation of ad platforms. Metahas the Ad Library. Google has its Ads Transparency Center. TikTok has the Creative Center. According to Similarweb, ChatGPT, AI Mode and AI Overview have none of that, neither for competitors nor, frequently, for an advertiser's own team.

That last point is the sharper one. A media buyer running Search, Shopping or Performance Max campaigns can end up serving inside a Google AI answer without any line in a report telling them so. Google Ads reporting does not segment performance for ads appearing within AI summaries, classifying them instead as Top Ads alongside other premium placements. Advertisers can neither target those positions directly nor opt out of them.

According to Similarweb, advertisers participating in Google's broader digital advertising programs are blind to ad placements in Google AI answers, most do not know whether their own ads are running in that context, and Google's visibility tools do not answer the question.

The mechanics differ on the OpenAI side. According to Similarweb, ads in ChatGPT surface only when intent exists and has been clarified from the conversation, with the outcome determined by the combination of ad creative, copy and landing page. There is no keyword to bid on and no placement report to read. Rather than leaving buyers to guess, the company states, AI Ads gives a direct look at what is winning placement.

The product ships with an AI Ads Gallery, described by Similarweb as a view of which ads are running and the context in which they appear. Insights are drawn from real user panel conversations rather than synthetic prompts, a methodological distinction the company puts at the centre of its pitch. The stated roadmap adds advertiser share of voice, ad categories and conversational intent in later releases.

Harel Amir, General Manager and Head of Product for Similarweb Ad Intelligence, framed the timing in the announcement. "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. Until today, nobody could see it at all," Amir said.

An advertiser perspective appears in the same release. Jonathan Bar Vardi, Head of Strategy at Natural Intelligence, a company Similarweb describes as a data-driven marketing specialist that has become a top buyer of ChatGPT ads, said: "When we started advertising on ChatGPT, we were flying blind, no visibility into who else was in the auction or what was working." He added: "Similarweb changed that. It shows what other advertisers don't know: who's spending, where they're appearing, and what's driving performance."

Both quoted parties have a commercial stake in the finding. Similarweb sells the measurement product that addresses the gap it documents, and Natural Intelligence is a buyer in the channel being measured. The underlying numbers are Similarweb estimates produced by its own panel methodology, and the company's standard disclaimer states that its data consists of or includes estimated metrics generated by proprietary algorithms, for which it accepts no responsibility for accuracy.

How the three numbers compare with other measurements

The 26% figure is not new to readers who follow the channel. Similarweb's 2026 Generative AI Landscape report, published on July 22, 2026, put ChatGPT ad penetration at 26% of US desktop chats during June 2026, against 14% in May. That report also recorded a click-through rate of 0.50%, found that 66.3% of ChatGPT ads appear after the second prompt or later, and measured AI Overviews on 43% of US searches. Today's release restates the ad penetration figure without the geographic qualifier attached to it in July.

Competing measurements of the same surfaces diverge sharply. Adthena recorded a 0.00 percent ChatGPT ad frequency across 169,560 United Kingdom result scrapes during June 2026, and zero ChatGPT ad items among 29,237 catalogued across all markets, while measuring AI Overviews on 23 percent of UK searches and 18 percent of US searches in the same month. Weeks later the same firm counted 7,378 distinct ChatGPT advertisers in the week of July 13 to 20 and reported United States ChatGPT ad frequency of 24.7% against 5.8% for Google AI Mode, a set of figures the company has not reconciled publicly with its own June report.

The AI Mode comparison deserves care. Similarweb's "nearly 30%" applies to ad-eligible queries, a narrower denominator than all queries. Adthena's 5.8% figure is not stated on the same basis. Two vendors measuring the same surface with different panels, different denominators and different collection methods will not produce comparable numbers, and no third-party measurement standard exists for any of these environments.

That is the structural condition the new dataset enters. Measurement of AI advertising is currently being built in public by the vendors who sell it, and the buy side has no independent arbiter to check the vendors against each other.

Why the gap opened

The surfaces arrived faster than the reporting around them.

Google introduced ads within AI Overviews for mobile users in the United States in October 2024. Adthena first detected them running in production on November 24, 2025, finding 13 instances across 25,000 monitored results pages, a frequency of 0.052%. Weeks later, Google expanded AI Overview ads to 11 further countries on December 19, 2025 without a formal announcement, a change advertisers discovered through documentation updates. Shopping ads in AI Mode followed on February 11, 2026.

OpenAI moved on a similar timetable. The company confirmed advertising plans for ChatGPT's free and Go tiers on January 16, 2026, launched the pilot on February 9, and crossed 100 million dollars in annualised revenue within six weeks. A self-serve Ads Manager with cost-per-click bidding opened to all US advertisers on May 5, 2026. The pilot reached the United Kingdom on June 6, and AdRoll opened access for a selected group of small, mid-market and business-to-business customers on August 11, 2026, six days before today's announcement.

Roughly eighteen months separate the first AI Overview ad tests from a channel with self-serve buying, conversion tracking, customer-list targeting and reseller access. The disclosure infrastructure did not travel at the same speed. Google's Ads Transparency Center, operating globally since 2023 and now carrying payer information for verified advertisers, was designed around a search results page, not around a generated answer that differs for every user.

Some platform-side reporting has begun to appear. Google introduced AI performance insights in Merchant Center on May 27, 2026, covering share of voice across AI Mode, AI Overviews and the Gemini app. That report addresses organic product visibility rather than paid placement, and share of voice there is determined by product data quality rather than by an auction.

What the panel method can and cannot show

Similarweb's stated differentiator is the source of its observations. According to the company, the insights derive from real user panel conversations rather than synthetic prompts, which it presents as a window into ads actually served to users.

The distinction matters technically. Synthetic prompting, the approach behind most AI visibility tooling, submits constructed queries to a platform and records what comes back. Those queries carry no conversation history, no prior ad exposure and no account context. ChatGPT ad selection, on Similarweb's own description of the mechanism, depends on intent clarified across a conversation. A single-turn synthetic prompt cannot reproduce the second-prompt-or-later position where most ChatGPT ads were observed to appear.

Panel observation has its own constraints. Coverage depends on panel composition, and the release does not state panel size, geographic distribution or device mix for the AI Ads dataset. Nor does it state pricing, availability date, or which Similarweb subscription tiers include the data. The published links point to the AI Ads product page and the broader Ad Intelligence page.

The roadmap items are equally undated. Share of voice, ad categories and conversational intent are described as anticipated in upcoming releases, without a timetable.

Where this sits in Similarweb's business

Ad Intelligence has been assembled through acquisition. Similarweb acquired Admetricks, a Chilean display advertising intelligence provider, in March 2024, and bought The Search Monitor on March 31, 2025, adding paid search monitoring, trademark enforcement and affiliate compliance. The AI-facing product line built on top of that base includes the GenAI Intelligence Toolkit released on July 28, 2025 and AI Studio, launched on February 10, 2026.

The financial backdrop is one of moderate growth. Similarweb reported first-quarter 2026 revenue of 73.9 million dollars, up 10%, alongside a planned CEO succession process targeting mid-2027, with full-year guidance of 307 to 315 million dollars. The Business Wire newsroom page carrying today's release also lists a second-quarter 2026 results announcement scheduled for August 12, 2026 before market open, and an undated release stating the company surpassed 300 million dollars in annual recurring revenue and signed two multi-year enterprise contracts worth approximately 47 million dollars in total contract value, recognised over three years and signed during the second quarter.

A dataset that becomes the reference point for a new channel is a durable commercial asset. That is the position Amir's quote stakes out, and it is the position the roadmap is designed to defend.

What it changes for buyers

For a media buyer, the practical situation before today was asymmetric in a specific way. Google and OpenAI hold complete records of every placement served. Advertisers hold partial records of their own activity and none of anyone else's. AI Overview placements do not appear as a separate line in Google Ads reporting at all.

A third-party panel does not close that asymmetry. It substitutes an estimate for an absence. Whether the estimate is good enough to allocate budget against depends on panel coverage that has not been disclosed, and on a methodology whose outputs already diverge from a competing vendor's by an order of magnitude in at least one market.

What has changed is that the buy side now has more than one estimate to work from, and the ability to compare them. The AI Ads Gallery, if the creative it surfaces is representative, gives categories of advertiser a view of competitor copy in a channel where ad copy conventions have been forming without a public archive to study.

There is a regulatory dimension that neither vendor addresses. The European Union's Digital Services Act obliges very large online platforms to maintain public repositories of advertisements served. How that obligation applies to advertising inside a generated answer, where the served creative is contextual to a private conversation, has not been tested. A commercial panel filling the gap is not a substitute for a statutory register, and its existence does not resolve the underlying question of who is required to publish what.

Timeline

Summary

Who: Similarweb Ltd., the Tel Aviv-headquartered digital intelligence company listed on the New York Stock Exchange under SMWB, announced the dataset. Harel Amir, General Manager and Head of Product for Similarweb Ad Intelligence, is quoted in the release, alongside Jonathan Bar Vardi, Head of Strategy at Natural Intelligence, an advertiser the company describes as a top buyer of ChatGPT ads.

What: AI Ads, an addition to Similarweb's Ad Intelligence line, providing competitive analysis of paid placements across ChatGPT, Google AI Mode and Google AI Overviews. It includes an AI Ads Gallery showing which ads are running and their context, and draws on real user panel conversations rather than synthetic prompts. Advertiser share of voice, ad categories and conversational intent are listed as future additions without dates.

When: August 17, 2026, distributed over Business Wire at 8:00 a.m. Eastern Daylight Time.

Where: Coverage is described as worldwide and available for specific markets. The three headline statistics apply to ChatGPT responses, ad-eligible Google AI Mode queries and Google searches, without a stated geography in the release itself, though the 26% ChatGPT figure was published in July 2026 as a US desktop measurement.

Why: ChatGPT, Google AI Mode and Google AI Overviews carry advertising but publish no ad library equivalent to Meta's Ad Library, Google's Ads Transparency Center or TikTok's Creative Center. Google Ads reporting does not separate AI Overview placements from other top positions, leaving advertisers without visibility into their own delivery, let alone competitors'. Similarweb sells the measurement product that addresses the gap it documents, and no independent standard exists to reconcile its estimates with those of competing vendors.