AINEWS
'Agent Ads' are FAQ style text blocks built for AI crawlers, not for human eyes. Ally reports a 3.5x lift from AI driven traffic; the platforms are still deciding whether to allow them.
Ally Bank and the Project Management Institute are now paying to place ads inside the stripped-down web pages that AI assistants like ChatGPT read before answering your question. The ads are not aimed at you. They are aimed at the bot.
The ad format has a coined name in the marketing trade: "Agent Ads." Each one is a short block of text, often written as a question-and-answer pair, formatted so the crawlers (the programs AI companies use to read the web) can ingest it cleanly and quote it back when a user asks a related question. The premise is that if a model is going to summarize a brand's products, the brand wants to be the answer the model is summarizing.
The most visible live test is a partnership between the adtech startup Mobian and the publisher Time. Time built text-only, crawler-optimized versions of its editorial pages so AI assistants could read them faster and more cheaply. Mobian then sold brands the chance to place ads on those pages, structured so an AI agent would pull the ad copy directly into its reply. Mobian's product page pitches the format as a new channel for "AI-mediated discovery" rather than a human-facing banner.
Ally's chief marketing officer, Andrea Brimmer, told Business Insider that consumers arriving at Ally from AI platforms were 3.5 times more likely to open an account than those coming from traditional search, and that AI-driven traffic to Ally is up roughly 9x year over year. Her framing: Ally would "rather be in the room helping shape what comes next than standing outside it." The Project Management Institute, another test partner, has not published traffic data; its CMO, Menaka Gopinath, said the appeal was that the ads were clearly labeled as ads, so consumers and the models reading the page could tell what was sponsored.
Mobian's CEO, Jonah Goodhart, told Business Insider that models are already influenced by advertising because they ingest the open web, and that a labeled ad slot beats letting a brand win the answer through search-engine optimization alone. On X, marketer Michael Bishop framed the milestone in the same terms, posting that the first brand to buy an Agent Ad "will own the answer" for years.
BSCI, a marketing accountability firm, raised the comparison to Bell Pottinger, the British PR firm that collapsed after running a covert influence campaign in South Africa. BSCI's Bernard Urban, also to Business Insider: "When the backlash comes, and it will come fast and furious, do you want to be the poster child for that backlash?" The deeper concern, echoed in the X thread and in marketer WhatsApp groups cited in the reporting, is that anyone willing to pay for an Agent Ad could shape a model's recommendation in ways users never see, and that the line between advertising and manipulation will not hold once the practice scales.
ChatGPT, Gemini, Claude, Perplexity, and Grok all currently treat agent-visible text as ordinary page content, which means a labeled ad block is no different from a labeled banner on a human-facing page. If any of them decides to treat agent-targeted ad copy as spam, or to strip FAQ blocks that look like ad creative, the entire test collapses overnight. As of this month, none of them has said it will.
The brands in the test are betting that being early matters more than being safe. The platforms that actually serve the answers will decide whether that bet pays off.