The lawsuit and the licensing deal are the same negotiation

In July 2024, News Corp sent Perplexity a cease-and-desist letter. Nobody answered, according to News Corp. In October it sued on behalf of The Wall Street Journal and the New York Post, alleging that the answer engine copied and summarized "large quantities" of its articles, analyses and opinions without permission.

The complaint goes further than scraping. News Corp says Perplexity sometimes reproduced its work verbatim, and sometimes "falsely attributed facts or even invented news stories under News Corp's name." It asked the court to make Perplexity stop and to destroy any databases holding its work.

In the same breath, CEO Robert Thomson praised OpenAI for taking the other road and signing deals with News Corp and other publishers to use their content legally. The coverage also noted that Perplexity had already started paying some publishers, Time and Fortune among them.

So one AI company was in court, another was at the table, and the defendant was already writing checks to other publishers. Those are News Corp's claims, as Digital Watch reported them. But the pattern is plain: AI companies either pay publishers or end up in court.

The link used to pay the bill

For about twenty years the deal between publishers and search engines was never written down. Search sent readers. Readers saw ads and hit paywalls. The link was the unit of credit, and credit turned into money on the publisher's own page.

Answer engines broke that deal by answering the question. Perplexity, in the coverage's phrasing, pitched itself as a way to help users "skip the links." Pew measured what that does at Google. Across 68,879 searches by 900 US adults in March 2025, people who got an AI summary clicked a regular result in 8% of visits. People who didn't get one clicked in 15%. Links inside the summary got clicked in 1% of visits.

That's one search engine, one month, one panel, but it's enough to see the scale of the problem. The reader still gets the publisher's work, compressed into a paragraph (and sometimes paraphrased incorrectly). The publisher gets about half the visits it used to.

When a publisher sues, then, it's arguing about price. The work is still being used. What went missing is the mechanism that paid for the use. A lawsuit puts a number on that in public, and a licensing deal puts a number on it in private. Thomson suing one AI company while praising another for paying is one position, held consistently.

Your work has two prices now

Anyone who makes trusted content now owns work with two values. One is what it's worth to people reading it. The other is what it's worth to machines that train on it, retrieve it and cite it. The first is shrinking for a lot of publishers, for the reasons Pew measured. The second is being priced right now, in court filings and contract terms.

The second market rewards different things. The citation is becoming the unit of credit, and the early data on who gets cited is lopsided. Profound's analysis of 680 million citations found Wikipedia supplied nearly half of ChatGPT's top-ten sources and Reddit nearly half of Perplexity's. A roundup of vendor studies reports, citing enterprise agentic marketing platform Yext, that sites hosting original research or first-party content draw 4.31 times more citations per URL than listings do. Those are marketing vendors measuring their own market, so I'd trust the direction more than the multiplier. Machines cite what they can't find anywhere else.

That's the asset News Corp is defending. A Wall Street Journal story is worth something to a machine because the Journal reported it, ran it first, and can be held to account for it.

What you license is what you can prove

The invented-stories allegation changes what the case is about. Copying a story takes something from a publisher. Making one up and putting the publisher's name on it spends that publisher's credibility on work it never did.

The BBC found the same failure when it tested four AI assistants on 100 news questions in December 2024. Of the quotes the assistants attributed to BBC articles, 13% were altered or didn't appear in the article cited, and 19% of answers citing BBC content introduced factual errors. I wrote about where those errors land in 45% of the time, the AI gets the news wrong. They land on the name.

This is where the economics and the trust argument meet. A publisher can only license what it can prove it made. It can only sue over work it can show was its own, published at a time it can document, in a form it can match against what the machine produced. And it can only fight a fabrication if it can show the story never ran.

Provenance used to be a defensive tool, the thing you reached for when someone faked your photo. In this market it's inventory. A dated, signed, archived record of what you published is what a licensing deal gets written against, and it's the evidence a lawsuit gets built on.

If you're not News Corp

Most people producing trusted content will never sue an AI company or get a licensing call from one. The repricing reaches them anyway. If your value is attribution, and attribution is now handed out by machines that may or may not show your name, the work that holds up is the work you can prove is yours: original data, first reporting, a dated archive, a public record of what you said and when.

The publishers in court and the publishers on the payroll are making the same bet. Machines want the work they can't get anywhere else, and they'll pay for it only when its owner can prove it.

Trust you can't prove you own is trust nobody has to pay for.