Analyzing the News: The IMF's October forecast

Thirteen outlets covered the IMF's latest forecast for the world economy. Fíor read six of them as AI and seven as human.

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

The IMF published its October World Economic Outlook on 13 October, during its annual meetings in Bangkok. It expects the world economy to grow 3.1% this year, up from 3.0% in its July update, and 3.4% next year. Global inflation is forecast at 4.7% this year.

The report credits investment in AI with holding growth up. It also says that growth is narrow, with 32 economies accounting for 95% of it, and that a pullback in AI spending could hit markets. Energy prices and public debt are the other main risks.

The outlets were Reuters, AP, AFP, UPI, Anadolu Agency, Bloomberg, the Financial Times, CNBC, Sky News, Euronews, the Guardian, Al Jazeera and the South China Morning Post. That is five wire services, the business press, British and European broadcasters, and outlets based in Qatar and Hong Kong. The Bloomberg piece was scanned as republished by The Edge Singapore.

The result

Six read as AI: AFP, the South China Morning Post, Euronews, the Guardian, Anadolu Agency and Bloomberg.

Seven read as human: UPI, Al Jazeera, the Financial Times, Sky News, AP, Reuters and CNBC.

On the face of it, that is a narrow win for the human side. It is narrower than it looks. Four of the seven human results were close calls: Reuters, AP, CNBC and Sky News. Only UPI, Al Jazeera and the Financial Times read clearly as human. All six AI results were clear.

The three that read clearly as human

The Financial Times piece was analysis rather than a report of the day's numbers. It set the new forecast against decades of earlier IMF forecasts and asked what stronger demand means for central banks.

Al Jazeera went to one region. Its piece covered the effect of the war on the Middle East and Central Asia, country by country, from Kuwait and Iraq to Kyrgyzstan.

UPI ran a straight news report, but a detailed one, with forecasts for the US, Spain, France, Japan and India.

The common thread looks like specific detail beyond the headline figures. It does not hold up on its own. Euronews also led with country figures, for the eurozone, Germany, the UK and France, and it read as AI. The South China Morning Post narrowed in on China, with policy detail and comment from analysts, and read as AI too.

Wire copy did not lean one way either. AFP and Anadolu read as AI. Reuters, AP and UPI read as human, two of them narrowly.

Why forecast coverage reads like this

Every outlet worked from the same report, released at the same moment, with the same headline numbers. Most pieces carried the same 3.1% growth figure.

Forecast stories also have a fixed shape. The number, the change since last time, the risks, a line from the Fund. A reporter filing quickly and a model given the same report would land on much the same structure.

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

The IMF's warning was about concentration. Most of the world's growth now comes from a small number of economies and one kind of investment. The coverage had its own version of that: thirteen outlets, one report, and much the same numbers in almost every piece.

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.