
Time’s experiment in advertising to machines rather than people is a month old, and it has already produced a rival’s public rebuke. Since 28 July, the publisher has been selling “Agent Ads”: sponsored question-and-answer blocks that only AI crawlers ever see. Chief operating officer Mark Howard told The Media Copilot that the format served more than two million agent impressions in its first month.
The experiment matters because AI agents are becoming a buying audience in their own right. Assistants such as ChatGPT, Gemini and Perplexity now research and compare on a person’s behalf, and some can complete the purchase. Bots generated 53% of all web traffic in 2025, according to the Imperva Bad Bot Report published by Thales in April. Brands that spent a decade optimising pages for search engines now face a harder question: how to put accurate, favourable information in front of software that reads the page, keeps what it finds and never sees a banner. Time and Wayfair, as Marketing Brew reported on 8 September, have given two very different answers.
Time’s Agent Ads work by serving AI crawlers a different page from the one human readers see. Time keeps an allow list of about 70 approved agents and bots. When one of them requests an article, it is routed to a stripped-down markdown version, and that page carries a sponsored FAQ block with brand-verified facts, labelled as sponsored content from start to finish.
The ad-tech company Mobian builds the units. According to Digiday, which broke the story on 30 July, Mobian turns an advertiser’s brief into FAQ copy, converts it to a PDF for human review, and then tests how visible, favourable and accurate the brand appears in AI search answers. Advertisers can target by Time franchise or by date range. Each markdown page carries one agent ad.
The first two buyers were Ally Bank and the Project Management Institute. Time has not disclosed pricing; Howard told Digiday only that it charges a premium. His case for the format rests on traffic. On most days, he said, Time now sees more bot traffic than human traffic, and Time draws more AI crawler requests than most of the roughly 7,000 publishers in TollBit’s network. “This is a growing traffic source, and therefore a growing source of inventory,” Howard told Digiday.
Howard also argues that agents hold on to what they read. Some Time articles, he told The Media Copilot, have produced more referrals from AI tools than citations, which he takes as evidence that agents store the content for later answers. Marketing Brew reported that Ally’s own data showed AI-referred customers converted at higher rates and arrived further along the sales funnel, though neither company has published the figures.
Perplexity called Time’s agent ads deceptive because the sponsored copy is shown to machines but not to the humans who ultimately read the answer. On 11 August, Perplexity’s chief communications officer, Jesse Dwyer, told Digiday that the company would block markdown advertising on Time.com from influencing its agents. He warned that publishers running “deceptive advertising like markdown ads” risk “a reputational downgrade” in Perplexity’s search index, including a lower trust score.
The dispute turns on where disclosure lives. Time labels the ad clearly on the markdown page, with opening and closing markers and a named sponsor. Howard has described the format as “the most transparent and cleanest way” to reach agents. But once an assistant paraphrases that copy into an answer, nothing obliges the label to travel with it. As Türkiye Today noted on 16 August, no regulation, industry standard or shared technical rule currently requires an AI assistant to flag that a fact reached it through a paid placement.
The row has not yet had a visible commercial cost. Howard told The Media Copilot he had seen no fall in Perplexity’s crawling of Time and that Perplexity was still surfacing the ads. Perplexity, for its part, said its search and security teams are working on broader measures against markdown ads. For marketing teams weighing a similar buy, that leaves an open risk: a placement that one answer engine rewards, another may actively penalise.
Wayfair is trying to reach AI shopping agents through its own product data and open protocols rather than paid placements. Chief technology officer Fiona Tan told Marketing Brew that the retailer is adding more product attributes so agents have fuller, more precise information to work with. Wayfair already uses AI to enrich and correct attribute details across tens of thousands of products, Digital Commerce 360 reported in May.
The second strand is infrastructure. Wayfair is one of five co-developers, alongside Shopify, Etsy, Target and Walmart, of the Universal Commerce Protocol (UCP), the open standard for agentic commerce that Google announced on 11 January 2026. UCP-powered checkout for Wayfair and Etsy went live in Google’s AI Mode and the Gemini app in the US on 11 February. Wayfair describes the standard as a way to keep the experience “seamless, accurate, and built on a trusted foundation”, whether a customer shops on its site or through a partner.
Tan was careful not to oversell it: agentic traffic, she told Marketing Brew, is still early. “We want to just make sure we’re keeping very close to how consumers are changing,” she said.
The same split applies to B2B software marketers, whose products agents also research and compare. Wayfair’s approach improves what agents find in data and pages the brand controls. Time’s approach buys space on pages the brand does not own, and one of the best-known answer engines has already moved against it. The Helm has reported separately that agents fall back to rival vendors when they cannot read B2B pricing, which makes owned, machine-readable information the lower-risk place to start. Mobian’s chief executive, Jonah Goodhart, told Digiday that only about 15% of brands currently run markdown pages of their own.
Time has not disclosed pricing for Agent Ads or named any advertisers beyond Ally Bank and the Project Management Institute.