People Inc. is developing custom APIs that let AI agents retrieve journalism as Markdown, with author details, publication and update dates, and other trust signals attached. The publisher is moving beyond the assumption that machines should collect reporting by visiting the same HTML pages designed for human readers.
The development, described by chief innovation officer Jon Roberts at Digiday’s Media Strategies London event, reflects a change in how publishers may distribute their work. The question is no longer just whether an AI system can find an article, but how that system receives current, attributed material and what control the publisher retains along the way.
Table of contents
Jump to each section:
- Why People Inc. wants to move beyond web crawling
- Markdown carries the story, but metadata carries context
- Time, Forbes and The Economist are testing alternatives
- What publishers and marketers should take from the shift
Why People Inc. wants to move beyond web crawling
In reporting published November 2, Roberts described the cost of sending full webpages to AI systems. Headers, navigation, advertising technology and recommended links may help operate a site, but much of that code adds little value when an agent needs the article itself.
About 90% of a typical HTML page is unnecessary material for an AI agent, Roberts estimated at Digiday’s Media Strategies London event. This is his estimate, not a measured industry-wide benchmark.
“Every token that needs to be processed is expensive,” Roberts told Digiday, explaining why the publisher wants to reduce browser-oriented page material in machine requests.
People Inc. has already experimented with different methods of providing content to partners. It began with webpage crawling, moved to Amazon S3 content distribution that allowed updates and retractions, and is now working toward custom APIs. Each step potentially gives partners a cleaner route to reporting while giving the publisher more explicit ways to manage what it sends.
The API initiative is still being developed. No commercial launch date, pricing, named AI customers or licensing terms were disclosed in the report, so it should not be treated as a finished marketplace for publisher data.
Markdown carries the story, but metadata carries context
Markdown is an efficient way to deliver readable article text without much of a webpage’s presentation layer. But the more consequential element may be the information attached to it: who wrote a report, when it first appeared, when it changed and whether an older version should still be used.
For a content-retrieval system, those fields can help preserve attribution and distinguish a fresh update from a stale copy. They do not guarantee that an AI assistant will cite the original publication or interpret the article correctly. Those outcomes depend on the receiving system and its policies.
The shift also creates a product-design problem for publishers. A web article is arranged for reading, navigation, advertising and subscriptions. A machine-facing endpoint needs decisions about identity, permissions, formats, version history and withdrawal of content. People Inc.’s move from crawling to controllable distribution makes that difference concrete.
Time, Forbes and The Economist are testing alternatives
Other publishing companies are testing their own approaches, although the initiatives are at different stages. They should not be mistaken for a single agreed standard for AI distribution.
| Publisher | Reported approach | What remains open |
|---|---|---|
| People Inc. | Custom APIs serving Markdown and article metadata | Partner terms and availability |
| Time | HTML-to-Markdown conversion and experiments selling ads to agent visitors | Commercial scale of agent readership |
| Forbes | Markdown and a Model Context Protocol layer for Vetted reviews | Broader rollout and partner adoption |
| The Economist | Experiments making some marketing and B2B material easier for agents to access | How much material and under what access rules |
The comparison shows the range of options, from cleaner content formatting to controlled programmatic feeds and new advertising tests. It does not show that any one model has already solved publisher monetization or licensing.
What publishers and marketers should take from the shift
The timing is not accidental. Automated browsers and agents are becoming a more visible category of traffic, making it harder for publishers to treat every bot request as either legitimate readership or unwanted scraping.
7,851% year-over-year growth in AI-agent and agentic-browser traffic was reported in HUMAN’s 2026 State of AI Traffic research, as cited by Digiday. This is traffic-category growth, not a People Inc. audience or revenue figure.
For publishers, the immediate operational questions concern which agents may access content, which updates they receive, how attribution travels with the article and whether an access model supports the economics of original reporting. A cleaner API does not settle payment or copyright questions on its own.
For marketing teams, the practical lesson is narrower than a demand to rebuild every article in Markdown. Keep author information and update dates clear, maintain accurate structured content, correct outdated material and examine where authorized machine access makes sense. These are useful foundations for AI-mediated discovery, but they should not be sold as a guaranteed shortcut to AI citations.
Key Takeaways
- People Inc. is developing custom APIs that provide journalism in Markdown together with attribution and freshness metadata.
- Time, Forbes and The Economist are testing other approaches, but no universal publishing model for AI agents has emerged.
- Controlled machine access creates new distribution choices without resolving publisher revenue, copyright or citation questions.
