Publishers start writing for AI agents as Axios and Every test a new media business model

Axios, Every and other publishers are reworking how they package news for a new audience: AI agents that summarize, sort and act on information before a human reader ever opens a homepage. The shift matters because outlets are no longer optimizing only for clicks and subscriptions; they are also trying to be legible to large language models and automated feeds that increasingly sit between readers and the web.

That change is already visible in products, partnerships and publishing workflows. According to WIRED on 2026-10-27, Axios has launched structured feeds for AI agents through Axios Direct, while Every has built article formats and interaction tools meant to be read, remixed or summarized by models such as Claude and ChatGPT. Behind both efforts is the same question: if AI systems are becoming the first stop for news discovery, how do publishers get paid — and how do they stay visible?

Why publishers are suddenly designing for machines

The simplest answer is that the traffic funnel is changing. Axios says the open web is weakening, search is shrinking and LLM use is rising, a combination that pushes publishers to think less about getting a human to click a story immediately and more about making sure their reporting appears inside an AI-generated answer at all.

WIRED’s reporting describes a market in which publishers are increasingly building for AI agents rather than only for human readers. That includes direct deals that let AI platforms surface publisher content, as well as technical tweaks aimed at making stories easier for crawlers and models to parse.

Axios chief revenue officer Jacquelyn Cameron framed the shift plainly: “We believe that the media ecosystem is changing”. The line captures the strategic bet now being made across the industry: if readers are arriving through chatbots, automated briefings or agentic summaries, then the publisher needs a product that can travel through those systems without losing its value.

What is Axios Direct?

Axios Direct is the company’s most explicit attempt yet to build for AI agents first and full articles second. Announced at the Digiday Publishing Summit in September, the product uses structured feeds rather than a standard story page, giving automated systems a cleaner way to ingest the information Axios publishes.

That structure is central to the pitch. Instead of asking an agent to crawl a long-form story and infer the essentials, Axios is providing feeds that package the information in a more machine-friendly format. The company says its short, structured style already performs well for AI crawlers because crawlers prefer quick access to the main points.

The first Axios Direct feeds are aimed at corporate clients, not casual readers. One feed gives Axios Pro Deals agents access to M&A coverage. Another is intended to help train in-house LLMs and is planned for next year. Axios also plans a consumer-facing Direct feed next year, extending the model beyond enterprise use cases.

By the numbers

How is Every approaching the same problem?

Every is taking a more experimental route, but it is wrestling with the same underlying shift. The company has built article experiences that invite AI interaction, including buttons for reading some guides with Claude or ChatGPT. It launched an extendable article format two years ago, pairing a reporting database with a chatbot so readers — or agents — can move between narrative, structured information and machine-assisted interpretation.

That approach goes beyond simple distribution. Every is trying to turn the article itself into a flexible object that an AI system can work with more easily than a traditional page of prose. The company’s product strategy also suggests a belief that journalism may not have to remain a fixed end product; it can become a source file for other formats.

Dan Shipper, Every’s founder, hinted at that ambition when he said, “People are gonna do crazy, crazy stuff with this”. Every’s recent work makes that less like hype and more like a product roadmap: the company published an 8,000-word Codex piece that agents can reformat into an article, podcast or video.

Every is also still small enough to move quickly. WIRED noted the company is six years old and announced a $2 million funding round last year. Its annual subscription costs $288, which places it in a very different market than Axios Pro Deals, but the broader experiment is similar: make the journalism more usable inside AI systems and see whether that creates new demand.

Why it matters

The stakes are bigger than a handful of publisher product launches. If AI agents become a meaningful layer in news consumption, they could reshape where audiences discover journalism, how publishers measure value and which content formats win distribution. That affects not only traffic but also brand awareness, subscription funnels and the ability to monetize premium reporting.

For readers, the change could mean a more mediated experience of the news. Instead of visiting several publisher sites, a person may receive an AI-generated briefing that stitches together headlines, summaries and action items. WIRED described one such use case through Michelle Seo, who gets a daily briefing from the Muse AI agent Chattiana that compiles agenda items, to-dos and news links from newsletters.

That kind of behavior points to a broader shift: news is no longer just consumed in the form of articles. It may be filtered into briefs, task lists and prompts, with the underlying publisher content invisible unless the source is surfaced inside the model. That is why discoverability — not just readership — is becoming a strategic issue.

There is also a direct business implication. If AI systems can satisfy more of the user’s information need without a click, publishers may gain visibility but lose referral traffic. That is the tension at the center of every AI-agent strategy. Revenue may follow only if the publisher can persuade platforms, enterprises or consumers to pay for the structured feed itself.

Background: how publishers got here

This is not the first time media companies have been forced to adapt to an outside platform. Publishers spent years optimizing for search engines and social networks, shaping headlines, layouts and even article lengths around ranking systems they did not control. GEO — generative engine optimization — is the latest version of that game, focused on appearing in AI answers rather than search results.

There is also a technical reason the shift feels so abrupt. AI chatbots and large language models can summarize, rewrite and surface news content without sending users to publisher sites. That capability makes the publisher’s original article both more valuable and more vulnerable: more valuable because it supplies the source material, and more vulnerable because the model may deliver the gist without the visit.

Some of the early response has been technical rather than editorial. WIRED reported that AIOSEO launched LLM.txt support for websites to help AI discover and interpret content. In practical terms, this is part of a growing effort to make publisher sites more legible to machines that crawl, summarize and answer questions.

At the same time, the legal and commercial backdrop remains unsettled. The New York Times v OpenAI and Microsoft copyright case is still ongoing, a reminder that the industry’s future may be shaped as much by litigation as by product design. Publishers are pursuing deals that allow AI platforms to surface their content, but they are also evaluating how much control they are willing to cede.

The timeline so far

Who is trying what, and at what price?

The current field is uneven. Professor Rasmus Kleis Nielsen of the University of Copenhagen says the market is emerging but immature and uneven, which is a useful description of a sector where no single revenue model has yet become standard. Some publishers are pursuing platform deals, some are building their own AI-facing products and some are experimenting with direct reader tools.

The commercial logic differs by company:

• Axios is leaning into structured feeds for enterprise and, later, consumers.
• Every is building article formats and AI interaction tools that let models repackage its journalism.
• AI platforms are becoming distribution partners, even as legal disputes over training and reuse continue.

Price points also underline how early this market still is. Axios Pro Deals costs $2,499 a year, while Every’s annual subscription is $288. Those are not directly comparable products, but the gap shows how publishers are testing everything from high-value specialist information to lower-priced membership offerings in order to see where AI-assisted use creates demand.

What happens next?

The next phase will test whether AI-agent products can produce revenue without hollowing out audience relationships. That is the central open question. Publishers want the visibility that comes from appearing in AI answers, but they also need a business model that does not depend entirely on the goodwill of platforms or on a shrinking flow of direct clicks.

Several developments are worth watching:

1. Whether Axios can turn its planned consumer-facing Direct feed into a product readers will actually pay for.
2. Whether the in-house LLM feed becomes a meaningful enterprise offering or remains a niche tool.
3. Whether Every’s format experiments lead to measurable subscription growth or broader licensing opportunities.
4. Whether more publishers adopt GEO and machine-readable feeds as standard practice.
5. Whether the legal environment pushes companies toward licensing deals or slower, more cautious experimentation.

For now, publishers are still in the early stages of learning how to sell to both humans and machines. Axios is betting that structured feeds can become a product in their own right. Every is betting that journalism can be made fluid enough for AI systems to reshape without losing its substance. The industry has seen platform shifts before, but this one is different because the middleman may not be a social network or a search engine. It may be a model that never clicks at all.

“Time is money for crawlers”