Analyzing the News: The US economy grows 3.4%

Thirteen outlets covered the latest US growth figures. Fíor read two of them as AI, including one for the first time.

Fíor is a natural language classifier that predicts whether a piece of text was more likely written by a person or by AI.

On 29 October, the US Bureau of Economic Analysis reported that the economy grew at an annual rate of 3.4% in the third quarter, up from 2.2% in the second and faster than most economists expected. Consumer spending rose at a 3.8% rate, and business investment was strong, led by spending on computers and equipment for the AI build-out. Housing investment fell, and trade pulled the figure down.

The same morning, separate figures showed inflation still running well above the Federal Reserve's 2% target. The numbers arrived the day after the Fed held interest rates, and a week before the midterm elections. Most coverage landed on the same tension: an economy growing quickly on paper, and households still feeling squeezed by prices.

The outlets were Reuters, AP, ABC News, CNBC, CNN, NPR, The Hill, UPI, the Guardian, Axios, Bloomberg, Fortune and the New York Post. That is three wire services, American broadcasters and newspapers, the business press and the Guardian.

Three more were left out. The Washington Post's scan once again picked up an AI-written summary of reader comments. Barron's scan captured a redirect notice instead of the article, and MarketWatch's stopped at a subscription prompt. Bloomberg's page also carries an AI-written takeaways box. This time, the scan was checked to make sure it started with the reporter's own article rather than the box, so Bloomberg is counted.

The result

Eleven read as human: Reuters, AP, CNBC, CNN, NPR, The Hill, UPI, the Guardian, Bloomberg, Fortune and the New York Post.

Two read as AI: ABC News and Axios.

Both AI results were confident. For Axios, Brath's other label agreed that the article reads like AI. For ABC News, it was unsure how the piece reads. The New York Post was the least decisive human result, and still read clearly as human. None of the pages drew a writing-habit flag.

This is the thirteenth story scanned with Brath, Fíor's new model. ABC News has now read as AI in three of its last four appearances. The Guardian read as human, and has now read as AI in seven of the fifteen stories where it ran its own reporting.

Axios, for the first time

Axios has appeared in the series four times before under Brath, and read as human every time. This time it read clearly as AI, with both labels agreeing.

The obvious explanation is its house style. Axios writes in a fixed format, short blocks under the same set of headings, and that kind of structure is something AI writing tends to share. It does not hold up as the explanation, though. The format was the same in the four earlier stories, and Axios read as human in all of them.

So something in this article was different from the others, and Fíor cannot say what. That is worth saying plainly. A pattern needs more than one result before it means anything, in either direction.

Why economic data coverage reads like this

Data-release stories are built from a government table. Every outlet had the same headline number, the same breakdown and the same forecasts to compare it with, and most published within minutes of the release.

What they did with it varied more than usual. Fortune argued the headline figure hid weakness in housing and a stall in September. NPR set the national numbers against what households are feeling. Bloomberg dug into falling savings. All three read as human.

Some of these pieces may have had AI help, with a first draft, a rewrite or a final polish. Or data coverage has become formulaic enough that a reporter and a model would write it the same way. Fíor cannot tell those apart on its own.

The 3.4% figure is an advance estimate. The Bureau of Economic Analysis will revise it on 25 November, and again after that, as fuller data comes in. First readings get revised. That is true of economic statistics, and of patterns in a classifier's results too.

What Fíor is

Fíor is a machine learning classifier, currently in beta and available as a browser extension. It gives a prediction, not a verdict. It cannot see who sat at the keyboard, and nothing here is a claim about any journalist or publication. Reading as AI and being written by AI are not the same thing.

Treat it as a signal, not proof, and check the pages you read yourself.

Try the browser extension at Fíor.

Analyzing the News is a series where we run one major story through Fíor and see what comes back.