Trump’s first 100 days battling the press, news outlets’ shift to podcast videos and more.
The evolution of AI content licensing deals
The Washington Post has become the latest major publisher to strike a licensing deal with OpenAI, joining a growing cohort that now spans more than 20 news organizations.
It’s part of a familiar pattern: every few months, OpenAI locks in another publisher to bolster its content pipeline. But the terms of these agreements seem to be quietly evolving — subtly shifting away from the explicit language around training data that defined earlier deals, and raising fresh questions about what these partnerships now mean.
The Washington Post’s agreement focuses on surfacing its content in response to news-related queries. “As part of this partnership, ChatGPT will display summaries, quotes, and links to original reporting from The Post in response to relevant questions,” reads the announcement on April 22 about the Post’s deal with OpenAI. In contrast, past deals with publishers like Axel Springer and Time – signed in December 2023 and June 2024 respectively – explicitly included provisions for training OpenAI’s LLM on their content.
The Guardian’s OpenAI deal, announced in February 2025, has similar wording to The Washington Post’s announcement and no mention of training data. A Guardian spokesperson declined to comment on the deal terms with OpenAI. The Washington Post did not respond to requests for comment.
These somewhat subtle shifts in the language of the terms could signal a broader change in the AI landscape, according to conversations with four media legal experts. It could signal a change in how AI content licensing deals are structured in the future, with more publishers potentially seeking agreements that prioritize attribution and prominence in AI search engines over rights for model training.
Another factor to consider: these AI companies have already trained their LLMs on vast amounts of content available on the web, according to Aaron Rubin, partner in the Strategic Transactions & Licensing group at law firm Gunderson Dettmer. And because AI companies are facing litigation from media companies claiming this was copyright infringement – such as The New York Times’ case against OpenAI – if AI companies were to continue to pay to license data for training purposes, it could be seen as “an implicit admission” that should have paid to license that data all along, and not scraped it for free, Rubin said.
“[AI companies] already have a trillion words that they’ve stolen. They don’t need the additional words that badly for training, but they want to have the updated content for answers [in their AI search engines],” said Bill Gross, founder of AI startup company ProRata.ai, which is building tech solutions to compensate publishers for content used by generative AI companies.
Both AI companies and publishers can benefit from this potential evolution, according to Rubin. AI companies get access to reliable, up-to-date news from trusted sources to answer questions about current events in their products, and publishers “can potentially fill a gap that they were scared would be missing with the way that these AI tools have been evolving. They were losing clicks and eyeballs and links back to their pages,” he said. Having better attribution in places like ChatGPT Search has the potential to drive more traffic to publishers’ sites. At least, that’s the hope.
“It has the potential to generate more money for publishers,” Rubin said. “Publishers are placing a bet that this is how people are going to interact with media going forward.”
Since last fall, OpenAI has challenged search giants like Google with its AI search engine, ChatGPT Search – and that effort depends on access to news content. When asked if the structure of OpenAI’s deals with publishers had changed, an OpenAI spokesperson pointed to the company’s launch of ChatGPT Search in October 2024, as well as improvements announced this week.
“We have a direct feed to our publisher partner’s content to display summaries, quotes and links attributed to original reporting in response to relevant questions,” the spokesperson said. “That’s a component of the deals. Post training helps increase the accuracy of responses related to a publishers’ content.” The spokesperson did not respond to further requests for comment.
How much publishers like The Washington Post stand to make from deals with OpenAI is unclear, especially as a different model may be emerging focused on ChatGPT Search. But the outlook for licensing deals between publishers and AI companies appears to be worsening. The value of these deals are “plummeting,” at least according to The Atlantic’s CEO Nicholas Thompson, who spoke at the Reuters Next event last December.
“There’s still a market for licensing content for training and that’s still important, but we will continue to see a focus on entering into deals that result in driving traffic to sites,” said John Monterubio, partner in the Advanced Media & Technology group at law firm Loeb & Loeb. “It’s going to be the new form of SEO marketing and ad buying — to appear higher in results when communicating with these [generative AI] tools.”
What we’ve heard
“We don’t have to worry about a somewhat false narrative of: cookies must go…. Then you can actually put all of this bandwidth and horsepower onto improving the current market, not worrying about a potential future problem that was in Google’s control all along.”
– Anonymous publishing exec on Google’s decision last week to continue using third-party cookies in Chrome.
Numbers to know
$50 million: The amount The Los Angeles Times lost in 2024.
50%: The percentage of U.S. adults who said AI will have a very or somewhat negative impact on the news people get in the U.S. over the next 20 years, according to a Pew Research Center study.
$100 million: The amount Spotify has paid podcast publishers and creators since January.
0.3%: The expected decline in media usage (both digital and traditional channels) in 2025, the first drop since 2009, according to PQ Media research.
What we’ve covered
AI lawsuits highlight publishers’ struggles to block the bots from scraping content
Ziff Davis’ recent lawsuit against OpenAI highlights the reality that publishers still have no reliable way to stop AI companies from scraping their content for free.
While tools like robots.txt files, paywalls and AI-blocking tags have emerged, many publishers admit it’s very difficult to enforce control across every bot, especially as some ignore standard protocols or mask their identities.
Read more here.
Who would buy Chrome?
The Google search antitrust trial could force Google to part ways with the Chrome browser.
If it did, OpenAI, Perplexity, Yahoo and DuckDuckGo could be some of the potential buyers.
Read more about the potential impact of a Chrome sell-off here.
TikTok is courting creators and agencies to take part in its livestreaming tools
TikTok is trying to prove the revenue potential of its livestreaming tools.
The social media platform says its creators are now collectively generating $10 million in revenue daily through livestreaming.
Read more about TikTok’s pitch here.
WTF are gray bots?
Generative AI crawlers and scrapers are being called “gray bots” by some to illustrate the blurred line between real and fake traffic.
These bots can impact analytics and steal content, and AI-driven ad impressions can harm clickthrough and conversion rates.
Read more about why gray bots are a risk for publishers here.
Is Facebook becoming a new revenue stream for publishers again?
Publishers have witnessed a recent year-over-year spike in Facebook referral traffic, and it’s — somewhat surprisingly — coinciding with an influx of revenue from Meta’s content monetization program.
Of the 10 publishers Digiday spoke to for this article, several are on track to make between six and seven figures this year from Meta’s latest content monetization program.
Read more about what publishers are getting from Facebook here.
What we’re reading
News outlets’ podcast video ambitions highlight audio format’s move into TV
News outlets like The New York Times and The Atlantic are putting more resources into producing video from popular podcast shows to tap into younger YouTube audiences, Vanity Fair reported.
Perplexity wants to collect data on users to sell personalized ads
Perplexity CEO Aravind Srinivas said Perplexity is building its own browser to collect user data and sell personalized ads, TechCrunch reported.
President Trump targets the press in first 100 days
President Trump targets traditional media companies in his first 100 days, using tactics such as banning outlets from covering White House events to launching investigations into major networks, Axios reported.
Semafor will test subscriptions
Semafor will “try out” subscriptions in “due time,” founder Justin Smith told the New York Magazine’s Intelligencer in a deep dive on the newsletter-focused news start-up company.
