News outlets are beginning to produce content aimed directly at artificial intelligence agents rather than human readers, according to a new report from WIRED. The shift marks a fundamental change in how media operates and who its stories serve, as publishers respond to what they see as a deteriorating open web, shrinking search traffic, and surging hyperpersonalization through large language models. Major outlets including Axios and Every are experimenting with formats ranging from feeds optimized for AI consumption to articles structured to let agents generate personalized versions for individual users.
Axios announced at the Digiday Publishing Summit in September that it would launch Axios Direct, a series of feeds designed specifically for AI agents. The first feed, which launched this fall, provides AI agents of Axios Pro Deals subscribers with merger and acquisition content, while a second feed planned for next year will help train in-house large language models. A third consumer-facing feed is slated for development in 2025, allowing individual subscribers to receive Axios reporting sent straight to their agent. The Direct feeds resemble RSS feeds—standardized text lines lacking visual appeal, since they aren't meant for human eyes. Meanwhile, tech media company Every takes a different approach by embedding prompts in articles that readers can copy into AI to test concepts, and offering "Read With Claude" or "Read With ChatGPT" buttons that pull up instructions for agents to summarize content. Every recently published an 8,000-word piece about Codex accompanied by a database of reporting that subscribers' agents can use to tailor new articles, podcasts, or videos suited to individual interests.
Axios chief revenue officer Jacquelyn Cameron says the company believes "the media ecosystem is changing," pointing to the open web deteriorating and search shrinking as LLMs become more important in news consumption. She describes the Direct feeds as a more efficient method for agents to access Axios content, cutting back on narrative to prioritize structured data similar to a Bloomberg terminal rather than traditional articles. At Every, founder Dan Shipper explains that certain article types include buttons allowing AI to start with a tight bulleted summary before offering to go deeper, letting writers pack stories with details without worrying about losing readers' attention since agents will extract the most relevant portions. According to the report, Axios chief operating officer Allison Murphy notes that the publisher's short-form, highly structured style has already made it a favorite for AI to pull from when responding to prompts, explaining that "time is money for crawlers" who don't want to go three quarters down a page to find the main point.
The report explains that publishers face uncertainty about whether they can actually profit from treating AI as an audience. Rasmus Kleis Nielsen, a professor of communication at the University of Copenhagen, says the market for publishing to agents remains "emerging" and "extremely uneven," with major questions about price points and how AI agents will interact with content in the future. Even if readers don't visit publisher homepages, brand visibility from AI crawling can be positive, similar to how publishers create content for social platforms like TikTok and Instagram that rarely drive viewers back to websites. Publications are struggling with how difficult it's become to monetize content when a chatbot can deliver answers directly, meaning people only want to pay for information that other outlets and LLMs don't have—at Every, that means detailed guidance on AI usage, while at Axios Pro Deals it's rapid access to market-moving information. An all-access Axios Pro Deals subscription runs $2,499 annually, while Every charges $288 per year.
Murphy frames the shift as one of acceptance, saying people are already using AI to absorb information and more will do so soon, adding that "this isn't the first time this has happened in media" since people already get news via social media, short-form video, and podcasts. Shipper views his company's annual subscription as covering both a human reader and their agents, reasoning that while a subscriber may not want to read everything each month, they may want their agent to know it so it can flag important items. The approach reflects what Murphy calls AI serving not as the audience itself but as a tool helping publishers reach their audience. For publishers navigating this transition, the question isn't whether to adapt to agent-driven consumption but how quickly they can retool their production without losing the editorial substance that justifies charging hundreds or thousands of dollars annually. The risk is that in optimizing for machine efficiency, outlets may strip away the narrative texture that gives journalism its persuasive power and makes readers trust it enough to pay.