Unique visitors to leading US online publishers fell by a quarter between 2023 and 2025, according to new data shared by trade body Digital Content Next.
This has led to the start of a “downward trajectory” for advertising revenue and overall publisher revenue last year, DCN said.
Google began rolling out AI summaries at the top of search results from May 2024 while ChatGPT arrived in November 2023.
Both companies, and other tech giants like Microsoft, train their AI tools on publisher content and continue to scrape in order to get up-to-date information in response to queries via RAG (retrieval-augmented generation). The tech companies have since signed some publisher deals for use of their content but say they do not need permission to scrape as it is “fair use”.
DCN said there was a 26% reduction in unique visitors to a survey sample of its members’ sites between 2023 and 2025, an 18% drop in total page views, and an 11% fall in total minutes spent.
DCN shared the survey data from its members in a document filed on Thursday in support of The New York Times and other news publishers including The New York Daily News and Ziff Davis in their ongoing copyright case against OpenAI and Microsoft.
The New York Times is a DCN member, along with the likes of The Washington Post, News Corp, Advance Local, Conde Nast, Financial Times, Fox News and The Guardian.
Last year DCN shared data from a survey of 19 member companies, including “major national newsrooms” and “global entertainment brands”, showing their referral traffic from Google search fell by 10% on average in May and June 2025 compared to the same period a year earlier.
In the past year, DCN said, that decline has continued. It repeated the survey across the same eight-week period in 2026 and saw an “even more” dramatic drop of 22% on average year on year.
The court filing noted: “The effect of AI tools on web traffic has only gotten worse over time, as the technology continues to develop at a rapid pace.”
It cited a study published this year by Carnegie Mellon University and the Indian School of Business which found that Google AI Overviews “reduce outbound organic clicks to publishers’ webpages by 39.8%”. DCN added that ChatGPT and Microsoft Copilot (formerly Bing Chat) “almost certainly have a similar effect”.
As a result, DCN said, both advertising revenue and total revenue for its members “began a downward trajectory” in 2025, each falling by 5% compared to the previous year. “RAG is thus already eating away at publishers’ businesses.”
“AI companies’ wholesale piracy of original content has already throttled publisher revenues and, if allowed to continue, could break the back of the free press on which society, including those same AI companies, depends,” DCN said.
“In this vicious cycle, Defendants are draining the life’s blood from news and content providers every time they copy original content and present it as tool output.”
Advertising revenue made up 84% of revenue for DCN members who took part in a survey last year, while subscriptions and other revenue streams made up 16%.
Although advertising revenue is the most directly hit by the fall in traffic, DCN warned other revenue streams are also threatened.
It cited a 2025 survey of New York Times subscribers which found that among those who used or paid for ChatGPT for news, 36% agreed the AI tool “means I no longer need news from the New York Times at all”.
DCN added: “A news executive from another newspaper testified that subscribers are ‘cancelling’ because publishers are ‘competing with a zero price on ChatGPT, on Copilot.'”
“Every subscriber lost to a substitutive AI tool represents the loss of a renewing revenue stream that would otherwise fund publishers’ creation of new content,” DCN added.
Overall, DCN said, the behaviour of AI tools summarising publishers’ original content “poses an existential threat to the publishing industry” and is creating a “doom loop” – a term also used by Microsoft itself.
It said: “In order to provide high-quality results, RAG depends, as training data, on access to the same high-quality content that RAG is throttling, which underscores the shortsightedness of Defendants’ conduct … And yet, Defendants continue to pursue a course of killing the goose that laid their golden eggs.”
DCN argued that the scraping of publisher content does not meet the threshold of fair use in US copyright law, as claimed by the tech platforms, because the “AI products result in substantial negative impacts on the digital publishing market”.
In its own amicus brief, the Open Markets Institute similarly argued that fair use cannot apply when AI products substitute and undermine the creation of original reporting. “When the original and the copy operate in the same market and are used by the audience for the same purpose, they are substitutes, not complements.”
The News/Media Association, whose members include The New York Times, Ziff Davis and other plaintiffs in the case, told the court: “Journalism is an expensive undertaking. Copyright protection is essential to enabling news organisations to recover the significant investments required to carry out the work of covering breaking news, worldwide developments and local events.”
The New York Times and other news publishers have applied for a summary judgment in their favour to head off a trial in the case, as have OpenAI and Microsoft.
The news publishers cited internal memos from the tech giants themselves which described the use of copyrighted content to build AI tools as “an astonishing theft of unprecedented proportions” and perhaps the “largest theft of labour in human history”.
Email pged@pressgazette.co.uk to point out mistakes, provide story tips or send in a letter for publication on our "Letters Page" blog
