AI agents are becoming a genuine audience for your work, and most creator content is effectively invisible to them. That is the practical lesson from a change underway at People Inc., the US publishing group that is building custom APIs so AI agents can read its journalism without wading through pages written for human eyes.
Jon Roberts, the company’s chief innovation officer, told delegates at Digiday’s Media Strategies London event that roughly 90% of an HTML page is material meant for a browser rather than an agent. The fix is to serve the article text in Markdown with the details that signal trust: who wrote it, when it was published, when it was updated. The urgency is easy to understand once you look at the traffic figures. Requests from AI agents and agentic browsers grew 7,851% year over year, according to HUMAN’s 2026 State of AI Traffic report.
| Key Takeaway | What it means for you |
|---|---|
| Machines are a second audience | Agents now read your content before a human does, so structure decides whether you get found or skipped. |
| Clean structure beats decoration | Markdown, clear headings and clean text survive machine reading; heavy page furniture does not. |
| Trust signals are metadata | Author, publish date and update date tell an agent whether your work is current and credible. |
| Ownership protects distribution | If your content lives only inside someone else’s platform, you cannot restructure how it is delivered. |
| Product thinking pays off | Treating your archive as a product, not a feed, opens revenue routes that platform ad shares never will. |
| The right build partner shortens the path | get in touch with Cre8rs when you want the technical side handled by a team that works with creators daily. |
Why AI agents are now part of your audience
The 90% problem in plain terms
Think about what an AI agent actually receives when it visits a creator page. It gets navigation menus, cookie notices, related post widgets, newsletter pop ups, tracking scripts and a wall of styling code. The article itself might be a small fraction of what arrives. Every one of those extra tokens costs money to process and adds noise that can bury the useful part.
That is the problem People Inc. set out to solve, and it is not a publisher-only problem. If a creator runs a blog, a recipe archive, a podcast transcript library or a membership site, the same inefficiency applies. An agent trying to answer a question about your work may give up, summarise badly, or attribute your ideas to whoever formatted them more clearly.
The opportunity is straightforward. When an assistant recommends a product, a guide or a specialist, it draws on sources it can parse confidently. Content that arrives clean, dated and attributed has an advantage over content that arrives as soup. The creators who treat that seriously will quietly collect referrals that never show up in a social analytics dashboard.
What People Inc. is actually doing
People Inc. did not begin with APIs. The group first opened its archive to partners through web crawling, then through Amazon S3 buckets that allowed it to push updates and retract content. Custom APIs are the next step, because crawling is an inefficient way for a machine to gather information and it hands control of accuracy to the crawler.
Roberts was blunt about the maths. Every token processed carries a cost, and dumping an entire HTML header on an agent is wasteful. The structured version strips that away and hands over the article in Markdown alongside the metadata that helps an agent judge whether the material is fresh and trustworthy. That is a packaging decision, not a content decision, and it is one any creator business can copy at its own scale.
What creators can take from a publisher’s strategy
Trust signals are the new metadata
Dates matter more than most creators assume. An agent trying to answer a question about current pricing, current tour dates or current availability will discard undated material, because it cannot tell whether the information still holds. The same applies to authorship. Work that names a real person, with a real role, is easier to cite than work published under a brand handle with no human attached.
There is a commercial edge here too. When your archive carries reliable dates and clear attribution, it becomes usable as a reference source rather than a one-off read. That is how content starts earning attention months after publication, and it is a large part of why what Cre8rs builds puts structure and metadata into the foundations rather than treating them as an afterthought.
Practical steps cost little. Publish real publish and update dates, and keep one clear heading hierarchy. Write introductions that state what the page contains within the first two sentences, and avoid burying the substance under three promotional banners. These are habits that help human readers first, which is why they are worth adopting regardless of what happens to agent traffic next year.
Distribution is shifting faster than most creators realise
The traffic numbers explain the urgency publishers feel. A 7,851% annual rise in agent requests is not a rounding error, and it arrives at the same moment referral traffic from conventional search is softening. Publishers including The Economist have been testing ways to make marketing and sales material readable by agents, while others convert pages into Markdown to serve automated visitors directly.
For creators the implication is blunt. If a growing share of discovery happens inside assistants and agentic browsers, then the places you publish matter less than whether your content can be read once it is found. A beautifully designed page that an agent cannot parse is a dead end. A plain, well structured archive is an asset that keeps working. The Verge has tracked how quickly these automated visitors have moved from curiosity to a measurable slice of traffic.
The ownership question underneath all of this
Why renters get squeezed
Notice what a publisher needs before it can serve structured content to AI partners: ownership of the archive and the technical freedom to change how it is delivered. A creator whose entire body of work lives inside a social platform has neither. The platform decides the format, the metadata, the access rules and the price.
That asymmetry is the real story. People Inc. could make this move because it owns its journalism and its infrastructure, which means it can send updates, retract material and shape what partners receive. A creator relying on someone else’s app cannot retract a post, cannot add a trust signal and cannot negotiate terms. Independence is not an abstract virtue, it is the ability to make a decision like this one on a Tuesday and have it live by Friday.
The same logic applies to monetisation. First-party relationships, whether memberships, direct sales or email lists, keep the audience addressable. When the terms of a platform change, an owned list still works. Sprout Social has documented how consistently brands and creators with direct audience relationships outperform those dependent on rented reach.
Building the archive you can actually control
Owning infrastructure used to mean hiring a developer and maintaining servers. That requirement has largely gone. A creator platform today can be a focused app, a members area, a searchable library or a booking tool, built once and owned outright, with content structured properly from the first day.
The work is less intimidating than the language suggests. Decide what your audience actually needs, then choose a small number of surfaces that serve it well. Make sure every page carries the text, dates and authorship that make it usable, and let a technical partner handle the parts that are not a good use of your time. Cre8rs.io exists for exactly that gap between a creator’s ambition and a creator’s calendar.
Preparing your content for a machine-read web
Small changes with a long tail
Start with your best performing archive and make it legible. Move the substance into the first paragraph rather than behind a scroll of styling. Keep transcripts with your video and audio so the words exist as text. Update old posts with a visible revision date when the facts change. Name the person behind the work and describe what they do.
None of this requires a rebuild, and all of it helps a human reader who is skimming on a phone in a queue. The overlap is the point. Content built for clarity serves people and machines at the same time, and the effort is a fraction of what a redesign costs. As the Digiday coverage of the People Inc. shift makes clear, the technical work is mostly discipline rather than invention.
Then treat the result as a product. A structured archive can support a digest, a searchable library, a members-only briefing or a referral engine, all without starting from zero. That is where the commercial return sits, and it compounds as the archive grows. The creators who began this work two years ago now hold libraries that would be expensive to recreate.
Frequently asked questions
Do I need an API to be readable by AI agents?
No. An API is one route, and it suits publishers with large archives and partner deals. For most creators, clean structure, clear headings, real dates and readable text do the same job at almost no cost.
Will structuring my content for machines hurt how it looks?
It should improve it. Semantic headings, earlier substance and visible dates help human readers first. Styling stays intact, because the structured version is served alongside the designed page rather than replacing it.
How much of my traffic will realistically come from agents?
Nobody can give you a precise number, and it varies by niche. The direction is clear though. Automated requests are growing fast enough that publishers are investing in this now, and creators who wait usually rebuild from a weaker position.
Does any of this matter if I only publish on social platforms?
It matters a great deal. On social platforms you cannot control metadata, formatting or access rules, so you are entirely dependent on how that platform decides to present you. An owned site gives you the controls that make the rest possible.
Where should a creator start if the archive is already messy?
Pick the twenty pages or episodes that genuinely earn attention and fix those first. Add dates, tighten the introductions, keep transcripts as text. A focused clean up on your best work delivers more than a full audit of everything you have ever made.
Can I keep my personality if everything becomes structured text?
Yes, and there is no trade off. Structure is about where the words sit and what data travels with them. Voice, humour and opinion live in the sentences, and those survive any format change intact.
How does Cre8rs.io fit into this?
The team builds the apps, platforms and digital products that give creators owned infrastructure. That includes the kind of structured content and metadata work covered here, handled by people who understand creator businesses rather than generic web projects. You can start a conversation about what your archive could support.
Own the format, not just the follower count
The People Inc. story looks like a publisher problem, but the logic travels. Content that machines can read and credit is content that keeps working after the initial launch, and the only way to control how yours is packaged is to own the place it lives. That is a build decision, and the earlier you make it, the more of your archive benefits.
You do not have to rebuild everything this month. Pick one surface, structure it well, and give it a home you control. If you want a partner for the technical side, the team at Cre8rs works with creators on exactly this kind of project, quietly and without a sales pitch.
Source: digiday.com