The Related Press has quietly retooled for the AI period – structuring its archive so enterprise LLMs can reliably floor, cite and pay for many years of reporting.
The information group has spent the final 9 to 12 months making its tens of thousands and thousands of content material belongings throughout textual content, video, pictures and audio codecs, machine-readable for LLMs to assimilate simply.
The Economist and The Monetary Instances have already clocked the income alternative offered by the growth in enterprise retrieval-augmented technology (RAG) demand, and have been licensing their content material archives accordingly. In the meantime, Dow Jones’ Factiva unit — which has a community of 30,000 publishers — now operates an AI-licensed content material market for enterprises.
It’s a method that ought to safe AP’s future as an info knowledge repository for the AI period, and widen its buyer base (traditionally information organizations) to incorporate extra enterprise shoppers by assembly their AI wants, AP world chief income officer Kristin Heitmann advised Digiday.
Heitmann stated AP intelligence providing is aimed primarily at enterprises creating their very own LLMs or proprietary AI programs that require structured, verified knowledge. Earlier this yr AP launched on the Snowflake Market – a cloud-based trade the place corporations like AP can license structured, rights-cleared knowledge on to enterprises constructing AI programs.
Since doing so, AP has licensed its content material to Snowflake’s finance shoppers, a lot of whom are constructing inside AI instruments and want entry to reliable, rights-cleared info, added Heitmann.
“This providing actually covers not simply finance corporations, however corporations seeking to monitor their provide chain, taking a look at disaster and operations administration, environmental and regulatory consciousness. I really feel like there are virtually limitless use instances,” she stated.
Turning the AP’s whole archive into AI-readable gas is a quiet moonshot: it means 5 a long time of verified reporting are actually cleanly digitized, tagged, rights-cleared and served through APIs so enterprise RAG programs can belief, cite and pay for it.
There may be scant publicly obtainable knowledge to indicate proof of the dimensions of the demand from the enterprise sector, however to publishers it’s clear there may be an enterprise RAG growth that’s opening up a brand new income lane for them.
A Deloitte report printed final week confirmed a big share of massive corporations aren’t simply utilizing off-the-shelf AI instruments (like ChatGPT or Gemini) – they’re actively constructing or customizing their very own LLM fashions and RAG programs. The AI ROI examine of practically 2,000 massive private-sector corporations throughout 14 nations in Europe and the Center East discovered that the majority companies are mixing inside and exterior AI instruments. About 38% take a hybrid method, combining in-house programs with third-party instruments, whereas 32% rely primarily on vendor-built options for velocity and scale and 24% plan to concentrate on creating their very own in-house AI capabilities, per the examine.
With 230 areas in practically 100 nations, AP produces 5,000 particular person items of distinct content material on daily basis. “The worth of that eyewitness journalism in right now’s AI info age is extra necessary, and albeit extra useful, than ever earlier than,” stated Heitmann. “As a result of the AP has produced content material in a number of languages, codecs – textual content, pictures, video and audio – and has the related metadata related to it, signifies that our capacity to achieve the house could be very vital and that’s why we’re investing right here to make sure that we will thrive and survive within the new AI info age.”
AP started its AI licensing journey in 2023 when it struck a licensing cope with OpenAI. It signed an settlement with Google to license AP content material to Gemini firstly of 2025. And this week it revealed it’s additionally considered one of a handful of Microsoft’s writer companions of the tech firm’s AI content material market, together with Folks Inc. and USA Right this moment Co. (rebranded from Gannett this week).
Heitmann stated the content material market remains to be on the experimental section however that it’s important that the AP has a “seat on the desk” in its growth. Her feedback are in sync with Folks Inc.’s chief innovation officer Jon Roberts, who has stated that it’s going to assist Microsoft construct from the “floor flooring.’
Talking at The IAB Tech Lab’s Worldwide Summit in London yesterday, Folks Inc.’s Roberts likened the present situation the place AI crawlers steal creator content material with out permission to the advert tech tax usually taken by distributors within the digital promoting provide chain. Besides on this case, AI corporations that scrape content material with out offering remuneration to publishers are like DSPs taking a “one hundred pc take fee” as a substitute of the “rational” 5 p.c {that a} DSP ought to take.
Roberts stated that in 2024, traders and technologists believed AI would make publishers out of date — that machines might take up the world’s info and churn out content material with out ever needing journalists. That phantasm cracked in 2025, when China’s DeepSeek mannequin surprised the U.S. by matching main programs at a fraction of the price, and the open-source group replicated it inside hours. It grew to become clear that constructing highly effective AI required dependable, high-quality knowledge, he confused. In that mild, the worth of verified journalism appears to be like far much less replaceable than Silicon Valley as soon as assumed.
“The most effective AI has one of the best inputs and generates one of the best outputs. That is the inverse of rubbish in, rubbish out,” stated Roberts on the summit. “Which means that sooner or later, we want extra good high quality, prime quality content material, not much less. We’re going to remodel your complete multi-trillion-dollar world economic system based mostly on AI, and AI essentially wants nice info to be nice. Now that is attention-grabbing as a result of it truly places the incentives of the AI corporations and the publishers on the identical aspect of the desk, whereas, I believe, for the final couple of years…these had been two separate teams standing on reverse ends of the desk.”
