AINEWS
Ad-measurement firm Guideline claims a view of AI-assistant ad spend. Its data isn't public.
Perplexity's ad business is disputed. ChatGPT and Gemini are still building theirs. The first outside measurement is a vendor with no public data.
Perplexity's ad business is disputed. ChatGPT and Gemini are still building theirs. The first outside measurement is a vendor with no public data.
Guideline, an ad-measurement vendor, told the market on July 7 that it can now show how much advertisers are spending inside AI assistants like ChatGPT and Perplexity, and how that spend is performing, using transaction-level data rather than the platforms' own numbers. The claim positions Guideline as the first independent view into a market the platforms themselves are still building, but the data behind it has not been made public.
The release, a corrected version of a company press statement, says the vendor's Ad Intelligence product now covers advertising activity across AI platforms inside a dataset that the company says represents roughly $200 billion in annual media investment across 65 countries. The $200 billion figure is Guideline's own characterization of its coverage scope, not an audited industry total, and the AI-platform slice of that data is not broken out in the release.
Perplexity's ad business is contested in the same window. Trade press reported that Perplexity tested sponsored ads with U.S. users as a new revenue line, while a separate analyst write-up argues the company has moved away from advertising in 2026. Google used its Marketing Live 2026 event to discuss Gemini-side advertising, and Adweek reported on the rollout. OpenAI has disclosed that its ad pilot exceeded $100 million in annualized revenue within six weeks of launch, per Reuters and CNBC, though independent transaction-level data on ad performance across AI platforms remains unavailable.
A single vendor is now asking the market to take its transaction-level numbers as the closest thing to independent verification of AI-platform ad performance. The release does not name any client, agency, or platform that has validated the methodology, does not disclose how transactions are captured, deduplicated, or currency-normalized, and does not define what counts as an "AI platform" ad in a market where the platforms themselves are still negotiating what their ad products are. Industry coverage of the announcement has so far followed Guideline's framing rather than tested it.
The release leans on the words "verified" and "independent" as positioning rather than as findings. The corrected statement appends additional details at the end that emphasize Guideline's claim to be the first to offer a true picture of the AI-platform slice, language that does work the underlying numbers do not yet do. The rest of the ad-measurement industry has not produced a competing claim, and the platforms themselves are not disclosing transaction-level data.
The platforms control the auction, the targeting, and the conversion data, and they have an incentive to claim their ad businesses are scaling. The vendor controls none of those inputs and has not yet shown its work. The practical question for advertisers, agencies, and the institutional investors the release is addressed to is whether a single vendor's dataset, built without platform cooperation, can outperform the platforms' own numbers.
Three things would change the picture. A public release of the AI-platform slice of Guideline's dataset would shift the story from a vendor announcement into a market reading; the absence of one would leave it as positioning. A response from OpenAI, Perplexity, or Google on the methodology would do the same. So would the appearance of a competing independent measurement product, which would either confirm that the AI-ad market is large enough to support more than one scorekeeper, or expose how thin the underlying data really is.
Until those land, the AI-assistant ad market is being narrated by the platforms that sell the ads and the vendors that claim to measure them. The independent verification the press release promises is not yet something anyone outside those two groups has been able to test.