July 7, 2026
Crystal Ball? More Like Roast Ball
Is The Economist Always Wrong?
The magazine graded its own crystal ball with AI, and readers are absolutely roasting it
TLDR: The Economist used artificial intelligence to score decades of its own predictions and concluded it does okay overall, despite some huge misses on oil. Readers were far less impressed, saying the test dodged the real issue: getting the big calls wrong matters more than padding the score with small wins.
The Economist tried a bold bit of self-checking: it fed around 7,000 old editorials into artificial intelligence and asked the machine to judge which ones made real predictions and how often those calls came true. The magazine says the results are mostly reassuring. Safe, mainstream predictions did fine; wild swings did worse. It points to wins like calling Vladimir Putin’s long grip on Russia and spotting Bitcoin’s staying power early, while admitting some truly painful misses on oil prices and demand.
But in the comments, readers were not handing out gold stars. The biggest mood was: nice experiment, wrong test. One of the sharpest reactions argued that the real issue is not whether The Economist gets lots of small things right, but whether it blows the big, important calls. Another commenter delivered the kind of line that belongs on a T-shirt: they said the magazine "asked the wrong question and used the wrong tool to get a questionable answer"—which is about as close to a public flogging as finance-comment-section humor gets.
There was also side-drama: one user complained about posting paywalled content, another pointed out this was a repost, and one blunt critic declared The Economist is "wrong more often than random chance." Ouch. So yes, the article is about forecasting—but the real spectacle is readers treating the publication’s AI report card like a live comedy roast.
Key Points
- •The Economist says it analyzed roughly 7,000 editorials from this century with GPT-5.5 to evaluate its forecasting record.
- •About 1,400 editorials were identified as containing falsifiable claims about the future.
- •The AI scored each forecast for both contrarianism and eventual accuracy, with repeated runs averaged.
- •The article says forecasts aligned closely with conventional wisdom were generally more accurate, while highly contrarian forecasts tended to be less accurate.
- •Examples cited include incorrect oil-market forecasts, a correct 2007 call on Vladimir Putin’s continued rule in Russia, and a 2013 prediction that Bitcoin would endure.