The country most wary of AI is building into the newsroom anyway

By Mersiha Ajdinovic

Picture: Pexels/Jakub Zerdzicki

Some of the newsrooms furthest ahead with AI, and some of the audiences most suspicious of it, turn out to be in the same place. In New York the papers are among the most advanced users of AI in journalism, and their readers among the most sceptical of it. That gap is what I went to understand.

I am a Swedish journalist. I was an LSE Polis JournalismFonden Fellow last year. I spent time in London newsrooms asking how AI was changing journalism, and what that meant for the way news is made and shared on social media. The answers in the paper that I published were careful. AI helped with subtitles, transcription and first drafts, but editorial decisions stayed with humans, and experimentation was mostly voluntary. This summer I took the same questions across the Atlantic, to The New York Times and The Wall Street Journal.

The audience is more suspicious, not less

Start with the public rather than the technology. American unease about AI is well documented, with polling consistently finding more of the public worried than excited about it. Rubina Madan Fillion, Deputy Editorial Director of AI Initiatives at the Times, sees the same thing from inside the newsroom. She told me US readers are, if anything, more sceptical than the ones I was used to in the Nordics, and her own reading is that it sits within a wider decline in trust in institutions over the past decade, journalism included. This is not a public waiting to be delighted by clever automation. It is one inclined to suspect it. Even inside the team building these tools, she said, the mood is split, colleagues who are "equally AI skeptics and AI optimists."

And the newsrooms are building anyway

In exactly that climate, the building was everywhere. The Times has a dedicated AI team, regular training, and reporting Madan Fillion said would have been "impossible without the use of AI." When reporters worked through the Epstein files, the newsroom built a tool that searched millions of pages by meaning rather than keyword, so they could find which documents deserved a human eye. Another in-house tool read handwritten notes from Syrian prisons, pulled out the text, translated it and sorted it.

The Journal had gone further into daily use. Tess Jeffers, its Head of Newsroom AI and Data, told me the paper has an enterprise agreement with Gemini and reckoned that between 50 and 75 per cent of the newsroom used it in some capacity in an average month. AI was "slowly becoming a really necessary tool for everyone," she said, before drawing a firm line: that "doesn't mean it's part of our reporting or our writing process, which again, is still very much owned and operated by our journalists."

So there is the pattern. The audience most anxious about AI sits in the country where I found newsrooms most willing to build with it. The question is not whether that is contradictory, but how they hold both at once.

The willingness to test

When I mentioned this to LSE professor Charlie Beckett, who leads the Polis think tank and the JournalismAI project, he made a fair point: he had heard close to the opposite from other US editors. His read was that newsrooms doing well are simply less afraid of AI. I agree, and would put it plainly: two of the strongest, best-funded titles in the country are no stand-in for its whole press.

That is not to say testing is a luxury. Every newsroom needs to try, in some measure, and the ones that dare to are the ones that come through best. That was the argument I made in my London report too, that the newsrooms which will come through this are the ones willing to try rather than wait for certainty, and that an AI mistake, caught and corrected, is not worse than the human mistakes every newsroom already lives with. What differs is not whether you should test, but how easy it feels. Being as relaxed and well read about AI as the Times and the Journal were is simpler when the business is doing well. The willingness is open to everyone. The comfort is not evenly shared.

That is what I saw in New York. The Times builds tools and runs them through human review. The Journal set guidelines and training around its use so reporters could try things and answer for them. What sets them apart is not how big they are. It is that they test in the open and stand behind the result.

I should add that I saw London a year before New York, and a year is a long time in this field, so some of what looks like a difference between the two cities may be timing as much as place.

When in doubt, show the journalist

Being willing to test explains who builds. It does not, on its own, win over a suspicious reader. For that, both newsrooms fall back on the same instinct: the AI never gets the last word. Madan Fillion described their process as "a human expertise sandwich," with editors and reporters at the start, AI in the middle, and human judgement at the end. "Cheat Sheet's making zero decisions," she said of one internal tool. The AI could tell you which hundred documents in a pile of ten thousand were worth reading. It could not tell you what the story was.

That instinct is sharpest on social media, where a wary audience is quickest to sense what feels machine-made. Instead of handing the voice to AI, both newsrooms are pushing the human further forward. The Journal is making more "social media takes on stories," filming reporters talking about their own work, and Lydia Serota, its Deputy Head of Digital, described moving "more into the sort of person or sort of human-led approach." The Times has put journalists into short videos explaining how they got a story, popular enough that its app now has a scrollable video feed. The machine works in the background while the reporter steps into the frame.

Make them known its you

This is where it meets the older question, the one behind my London report: what happens to a newsroom's identity when so much of what fills a feed is shaped by the same handful of tools? Jeffers named her worry plainly: what happens when "our facts show up on these sites but our brand does not"? The Times was chasing a version of it internally, testing whether AI-generated text actually feels like the Times, scored by editors on accuracy, tone and clarity. Not just whether a machine can write something correct, but whether it still sounds like them.

Seen that way, the suspicion and the building were never really at odds. In a country that distrusts the tool, the newsrooms confident enough to bet on it are the same ones working hardest to keep a human face on what comes out. A wary audience is not a reason to hide the technology. It is the reason to put the journalist where you cannot miss them. AI is already in the newsroom. What these two show is that the more of it they use, the harder they work to keep you looking at the human who used it.