More people are using AI to navigate the news, and some publishers are beginning to create content for the agents that help them. The shift could change how reporting is packaged, how readers encounter it, and what they are willing to pay for.
Newsrooms are adapting to AI discovery
Publishers face a changing route between their work and their audience. Search is shrinking, personalized information is gaining ground, and large language models are becoming a more important way for people to consume news, according to Axios chief revenue officer Jacquelyn Cameron.
Many major publishers, including WIRED’s parent company Condé Nast, have signed agreements that allow AI platforms to surface their content. Others are working on generative engine optimization, or GEO, to improve the chance that their reporting appears in answers to AI prompts. AIOSEO, a WordPress SEO plug-in, launched a feature that supplies websites with LLM.txt files to help AI systems find and interpret their material.
Publishers are also weighing how to protect their work. The New York Times sued OpenAI and Microsoft in 2023, alleging copyright infringement over the use of its reporting to train models. The case is still ongoing.
Axios is preparing feeds for agents
Axios is developing a more direct way to deliver news to AI systems. At the Digiday Publishing Summit in September, the company announced Axios Direct, a set of feeds designed for AI agents. The format favors structured information over the narrative style of a conventional article, making it closer to a Bloomberg terminal than a collection of stories.
The first two feeds are aimed at corporate customers. One, launched this fall, gives AI agents used by Axios Pro Deals subscribers access to merger-and-acquisition coverage. Another, planned for next year, will provide material for companies to train their own large language models.
A third feed, also planned for next year, is intended for individual subscribers. It would send Axios reporting to a subscriber’s agent. The feeds are expected to resemble RSS: standardized lines of text meant for software to process, rather than a page designed for people to browse.
That business-first launch reflects the fact that companies may be more willing than individual readers to pay for early experiments that could provide a business advantage. Whether these feeds become a lasting product will depend on how customers use them and what value they see in direct access.
Every gives readers’ agents more material
Every takes a different approach. The six-year-old technology media and software company builds AI use into the reading experience, including prompts that let readers try ideas from an article for themselves. A story about a new model, for example, may include prompts readers can copy into AI to see how the model responds and evaluates input.
Some guides include buttons such as “Read With Claude” or “Read With ChatGPT.” These open a prompt that asks the AI to begin with a brief summary and offer more detail. Writers can include more information, with an agent helping each reader find the parts most relevant to them.
Every has also developed “extendable” articles. These pair a long-form story with a database of reporting, including interview recordings, news articles and YouTube videos. A reader can ask a chatbot for additional details drawn from that material.
A recent 8,000-word article about Codex takes the approach further. Subscribers’ agents can use its database to create different formats, such as a bulleted summary, a podcast, or a short film made from interview clips. The idea is to let readers choose how they receive the reporting, rather than relying only on the writer’s original presentation.
The business model is still uncertain
Publishing for agents remains an emerging and uneven market, says Rasmus Kleis Nielsen, a professor of communication at the University of Copenhagen. Questions about pricing and how agents will interact with content still need answers.
For publishers, visibility in AI responses may have value even when readers do not visit the original website. Nielsen compares that possibility with publishing on social platforms such as TikTok and Instagram, which may not send viewers back to a publisher’s site.
But visibility alone does not settle the revenue question. Publishers are looking for information that readers cannot easily get from other outlets or a chatbot. Every focuses on detailed guidance about AI and how to use it; Axios Pro Deals sells fast access to information that could move the market.
An all-access Axios Pro Deals subscription costs $2,499 a year, while an annual Every subscription costs $288. Every founder Dan Shipper views the latter as covering both the human subscriber and their agents. That approach treats an AI assistant as part of the audience for paid reporting, while leaving open how much readers will value that access.