Time Now Serves Ads Directly to AI Agents Reading Its Pages

In July, Time became the first publisher in the world to serve ads directly to AI agents, according to Michael Mraz, the magazine's General Manager and Chief Product Officer. "We strongly believe agent ads will be a big part of the future of marketing and we're proud to be early," Mraz told attendees at Contentstack's ContentCon conference in Chicago in October.

The move followed a blunt realization. More bots now read Time than humans do, Mraz said. "That raised a pretty fundamental question: Were we actually built for the audience we now have?" he said. The honest answer was no.

What was broken at Time?

Time's website was, in Mraz's word, "creaky." It ran on four different front ends, dozens of content templates, competing code bases and archive databases dating back to Windows XP. The publisher had last updated its content management system (the software that stores and publishes editorial content) in 2014. Any promise of fast adaptation had to clear that legacy first.

How did Time rebuild for bots and humans?

The replatforming took five months. Time chose Contentstack for its modular, headless architecture (a setup where the content store is decoupled from the page design), which makes future updates easier. The measurable payoff: time to first contentful paint (the moment a browser first renders visible content) fell 27% versus the old system, and conversion increased, Mraz said.

For machine readers, Time now publishes a Markdown version of every page, built exclusively for AI agents. The publisher embeds ads inside that Markdown layer and serves them directly to agents, using ad tech startup Mobian to create the format. Monetizing bot traffic is the bet; whether agents' downstream users ever see those ads is the open question Mraz did not address.

What else did the new system enable?

The rebuilt stack accelerated product launches. Time relaunched its newsletter business with five AI agents running it through what Mraz called an "AI harness" — a model layer containing the company's business rules and subscriber data. "The harness sits between Time's journalism, our business and the reader's inbox, and it enables our small cross-functional team to operate like a much larger one," he said. A new games section went from idea to launch in weeks instead of months.

Time had already embedded AI tools site-wide before the replatforming. A sitewide AI agent, powered by Time's archive, summarizes articles, reads them aloud and translates them into 13 languages, drawing only from Time's own journalism. "It extends the life and reach of every story we publish, making trusted journalism accessible to global audiences in new ways that feel immediate and personal," Mraz said. The magazine also built an AI-generated podcast: human editors pick the day's top stories, two AI voices host, and editors can approve, tweak or regenerate any episode before it goes live.

Time has outlasted TV news, the internet, social media and smartphones, and Mraz frames AI as the next disruption to absorb rather than a threat. "As user behavior changes, and it is changing fast, trusted journalism reaches fewer people every year if it doesn't take new forms," he said. Whether agent-served ads become revenue at scale, or a first-mover experiment, will depend on ad buyers following readers into the bots' layer.