The siren song of AI scale

Turning to AI to write stories will not save smaller outlets

Illustrated scene of sailors in a boat facing towering waves and rocky cliffs, with siren-like figures watching from above.

Well, this is deeply alarming. Charlotte Tobitt, for Press Gazette:

Researchers from the University of Maryland used AI detector Pangram to analyse 186,000 articles published online by 1,500 national and local newspaper brands between June and September 2025.
They concluded that “approximately 9% of newly-published articles are either partially or fully AI-generated”.

I think my full, professional and learned opinion on this is:

FFS.

Really? Really?

We’re rushing into using an early-stage technology, with major inherent flaws, to great our central product? As ever, I’m not arguing against the use of LLMs. I’m arguing that we need to be cautious about where and how we use them.

Killing trust in journalism, one prompt at a time

And, right now, the evidence is that using them to generate journalism is going to become a big problem really fast. Trust in journalism is at a deeply problematic low, if a stable one, according to the latest Digital News Report:

One more relatively positive sign is that overall trust in the news (40%) has remained stable for the third year in a row, even if it is still four points lower overall than it was at the height of the Coronavirus pandemic.

Still, 40% is not a passing grade, and doing anything to erode that trust will only hurt the profession. So, how do people feel about AI-written news?

Oh. Oh, dear.

12% of people are comfortable with AI-written news – and that number is falling. Well, doesn't this seem like the perfect way to make our trust problem much, much worse?

The seductive lure of AI scale

But, some details further down the Press Gazette piece might give us a sense of the context of the issue:

AI use was found to be “much higher” in smaller local outlets compared to large national newspapers. Some 1.7% of articles at papers with a circulation above 100,000 were labelled as AI-generated or mixed, versus 9.3% at smaller titles.

So, the issue is more prevalent at small outlets, possibly ones still looking at scale as their key business model, rather than audience relationship. But I’m not sure sacrificing the latter in the quest for the former will solve their problems.

Trust is precious. It’s hard won, and easy to lose. The FOMO of not pushing hard enough and fast enough will make some small publishers make terminal mistakes.

Don’t be one of them.