Analyzing the News: The AI sell-off, a second time

Eleven outlets covered another tech sell-off over the cost of AI. Fíor read three of them as AI, and most outlets landed the same way as last 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 30 October, US technology stocks fell sharply after results from Intel and Amazon. Intel's shares dropped about 29% after weak guidance, its worst day in decades. Amazon's cloud business kept growing, but investors focused on its spending, around $220 billion this year with more likely in 2027, and on its weaker cash flow. The Nasdaq fell 2.86%, its biggest one-day drop since June, and an index of chip stocks fell 7.2%.

It is the third time in a fortnight that worries about AI spending have hit the market. The question underneath has not changed: when will all this money pay back?

The outlets were Reuters, AP, the Guardian, CBS News, Business Insider, the Wall Street Journal, The National, Yahoo Finance, Quartz, NBC News and Forbes. That is two wire services, American and British newspapers and broadcasters, the business press, and The National from the United Arab Emirates. Only part of the Wall Street Journal's article was scanned. The Guardian and CBS News both carried some AP reporting alongside their own. Yahoo Finance's coverage was a live blog.

Two more pages were left out. ABC News ran AP's own report, so counting it would count AP twice. NDTV Profit's page carries an AI-written summary at the top, and it was not clear whether the scan picked that up with the article. As with the Washington Post in earlier pieces, a result that might mix the two is not counted.

The result

Eight read as human: Reuters, AP, the Guardian, the Wall Street Journal, The National, Quartz, NBC News and Forbes.

Three read as AI: CBS News, Business Insider and Yahoo Finance.

The National read as human, but not confidently. Business Insider was a split: Brath judged it likely written by AI, but said it reads like a person wrote it. Yahoo Finance was the clearest AI result, with both labels agreeing. None of the pages drew a writing-habit flag.

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

The same test, two weeks apart

This story offers something the series has not had before: a repeat. The last sell-off, on 22 October, was also scanned with Brath, and seven outlets appear in both.

Six of the seven landed the same way. Reuters, AP, the Wall Street Journal and Forbes read as human both times. Yahoo Finance's live blog and CBS News read as AI both times. AP's piece was again the short daily list of where each index closed, and again it read as human.

Business Insider was the one that moved, from human last time to AI this time. Even then, Brath said it still reads like a person wrote it.

Why market coverage reads like this

Market reports are among the most formulaic writing in journalism: index levels, percentages and the same few company names, every trading day. Most of this coverage still read as human. The pieces that took a view of their own read as human too. The Wall Street Journal asked whether the correction could turn into something bigger. The National quoted an investor who thought the market had overreacted.

Some of these pieces may have had AI help, with a first draft, a rewrite or a final polish. Or market 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.

Investors asked the same question for the third time in two weeks. Fíor asked its own question twice, of the same outlets, and mostly got the same answer. Getting the same answer twice does not prove it is the right one. It does suggest it is not random.

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.