The New York Times has begun testing its first generative AI tool for readers, rolling out AI-generated search summaries to a small subset of visitors in a move the paper hopes will fix its long-criticized search feature — and which has already unsettled some of its own journalists.
The test, first reported by Semafor media editor Max Tani on August 23, marks a quiet but symbolically significant shift for a publication that has positioned itself as the news industry's most prominent adversary of AI companies. For the Times' news arm, it is the first test of AI-generated text that reaches readers unmediated by a Times journalist or editor. For more context on this story, see our ongoing AI trends.
In recent weeks, the paper has quietly introduced a new search page for a small group of visitors. When those users type queries, the page responds with excerpts from and links to relevant Times stories — alongside AI-generated summaries of the paper's reporting on the topic.
What the Times Is Saying
In a statement to Semafor, Times spokesperson Graham James framed the feature as an experiment in discovery. "We are always testing new ways for our users to discover and engage with Times journalism," James said. "This experimental feature is an example of how we are using the latest technology to create a better search experience."
The search test is not the Times' only brush with generative AI in its products. Wirecutter, the Times' product-recommendation site, recently began testing an AI search feature called Wirecutter Finder, which surfaces AI-generated summaries and tips based on the site's reviews and recommendations. The paper has also experimented with other AI features in recent months across its non-news sections, keeping the newsroom itself largely at arm's length from generated text — until now.
A Newsroom Pushback Years in the Making
The new summaries have been met with skepticism inside the Times' own newsroom. According to Semafor, the media newsletter Breaker reported earlier this month that Times management had informed the paper's union it would test AI summaries in search — an announcement made during a contract bargaining session in July, which led to frustration within the editorial union. The guild says the Times is violating its own editorial standards around AI use.
The Times Guild has proposed a series of rules around AI, including the creation of an employee payment pool in which 22.5 percent of revenue from AI licensing deals for training would be split among unionized employees. It is also pushing for stipulations that any use of AI include human oversight — a clause that, as Semafor noted, would render the new search tool impractical.
Times Guild unit chair and senior staff editor Jim Luttrell did not mince words about the company's AI policies. "Management can change and even violate the company's policy in any way they want, as they are doing with the summarization tool," Luttrell said in a statement to Semafor. "Every time AI summaries have been deployed, they have made embarrassing mistakes. That's why we need strong guardrails on AI in the contract — to protect readers' trust in our journalism."
He added that the paper's current AI policies are "meaningless."
Catch-Up or Caution?
Semafor's Tani, speaking with several Times employees last week, assessed the move as a necessary if modest catch-up. AI-powered search and summarization have become standard features across websites, apps and inboxes — and, increasingly, at other news organizations. By that measure, a several-sentence summary built solely from the Times' own reporting is, as Tani wrote, not the most ambitious implementation of the technology. But it is a notable first step for a paper whose internal search tools have long lagged its otherwise best-in-class publishing products, according to the employees Semafor spoke with.
The tension is hard to miss. The Times remains locked in a closely watched copyright lawsuit against OpenAI and Microsoft over the use of its journalism to train AI models, a case that has made the paper's own handling of generative AI a subject of particular scrutiny both inside and outside the newsroom. The paper has also been among the most aggressive publishers in negotiating licensing terms for AI use of its content.
Why It Matters
For the broader media industry, the Times' experiment is a test case in how far legacy publishers can go with reader-facing generative AI before eroding the editorial guarantees that distinguish them from AI-native aggregators. The paper's answer so far is a deliberately narrow design: summaries grounded only in Times journalism, presented as search results rather than standalone articles, and clearly framed as experimental.
But the union dispute shows the harder constraint may be internal. The fight over the 22.5 percent licensing pool and the human-oversight clause is, at its core, a fight over who decides what counts as journalism when machines assemble the first draft. How the Times resolves it — at the bargaining table or in public — will be watched closely by every newsroom weighing the same trade-off between relevance and trust.
For readers, the immediate change is modest: if you are among the small group of users in the test, searching the Times' site may now return a tidy AI-written digest of its coverage at the top of the results. Whether that digest earns readers' trust the way a bylined story does is precisely the question the Times — and the industry — are now running live experiments to answer.
---
Stay Ahead of AIGet the latest AI news, analysis, and breakthroughs — all in one place.
Read more AI news →
