APIs for news

News organizations will face an AI reckoning in 2026 and a choice: They can keep blocking AI crawlers, suing AI companies, and lobbying for protectionist AI legislation — all the while making themselves invisible to the publics they serve on the next medium that matters — or they can figure out how to play along.

Unsurprisingly, I hold a number of likely unpopular opinions on the matter:

1. Journalists must address their civic, professional, even moral obligation to provide news, reporting, and information via AI. For — like it or not — AI is where more and more people will go for information. It is clear that competitors for attention — marketing and misinformation — are rushing to be included in the training and output of large language models. This study finds that “reputable sites forbid an average of 15.5 AI user agents, while misinformation sites prohibit fewer than one.” By blocking AI, journalism is abdicating control of society’s information ecosystem to pitchmen and propagandists.
2. AI no longer needs news. Major models are already trained and in the future will be trained with synthetic data. Next frontiers in AI development — see, for example, the work of Yann LeCun — will be built on world models and experience, not text and content.
3. Anyway, training models is fair use and transformative. This debate will not be fully adjudicated for some time, but the Anthropic decision makes it clear that media’s copyright fight against training is a tenuous strategy. Note well that the used books Anthropic legitimately acquired yielded no payment to authors or publishers, and if Anthropic had only bought one copy of each title in the purloined databases, it would not have been found liable and authors would have netted the royalties on just one book each.
4. AI is the new means of discovery online. I had a conversation with a news executive recently who, in one breath, boasted of cutting off all the AI bots save one (Google’s), and in the next asked how his sites will be discovered online. The answer: AI. Rich Skrenta, executive director of the Common Crawl Foundation, writes that if media brands block crawlers, AI models will not know to search for them, quote them, or link to them when users ask relevant questions. He advises publishers to replace SEO with AIO: optimization for AI. Ah, but you say, AI doesn’t link. No. This study compared the links in AI against search and found that ChatGPT displayed “a systemic and overwhelming bias towards Earned media (third-party, authoritative sources) over Brand owned and Social content, a stark contrast to Google’s more balanced mix.” The links are there. Whether users click on them is, as ever, another question. But if your links aren’t there, no one will click anyway.
5. Sorry, but AI will bring no bonanzas to news. The deals done so far — payments from big AI companies to various media moguls — are not at all about licensing of content. These have been lobbying expenses, to buy companies’ silence when it comes to PR, legislation, and litigation. In the future, AI companies may make some payments for particular content, but rather than blanket licensing, I predict these will be predicated on a per-use basis. Keep in mind that AI companies aren’t particularly profitable so far. Until they figure out their own business models — whether advertising (and we know what media’s own attention economy did to ruin the web) or commerce (agents advertising to agents), or subscriptions (a problematic model as long as AI remains commodified) — there won’t be much revenue to share.

In spite of it all, I predict that most in media will wake up one day in 2026 and realize that they have little choice but to be in AI, because that’s where the people formerly known as their audiences are and because all their competitors are there. So how should they accomplish that?

I propose that news organizations — especially the vast ecosystem of journalism left out of the moguls’ money grabs — create APIs (application programming interfaces) for news. That is, create the keys that would make their news accessible for AI so they can open discussions with the technology companies about the responsible use of their content, branding, links, and, yes, possibly payment.

The big AI companies aren’t going to negotiate with thousands of small outlets, so I suggest that local, independent, and academic publishers gather together to create joint APIs — not to cut off AI, but instead to make it easier to come to meaningful arrangements at scale. Here is an opportunity for small and independent media to leapfrog the moguls, providing news and information to AI’s users and getting attention, branding, and links while the moguls are still blocking, suing, and lobbying.

I’ll further recommend that publishers start crawling information on their own, gathering relevant data from and for their markets and making it available not as stories but as stores of knowledge available to AI. Those publishers can embed AI in their services to query that data (“show me every restaurant in the area that serves empanadas”) and to let readers ask for more information (see how Google Trends this year includes a button that lets out-of-it oldsters like me get caught up on all the hot stars whose names I don’t recognize). Wise publishers will do deals to make further use of AI, sharing raw reporting in NotebookLM so readers can query it; having AI transcribe; summarize, link to, and even make podcasts of local board meetings; translating text into audio and other languages, and so on.

I’m sorry that I have no predictions of riches to follow from AI. But AI is a reality media must reckon with — and now.