July 5, 2026
The forecast? Comment-section chaos
Is The Economist Always Wrong?
The magazine put AI on its own crystal ball — and readers say the answer is painfully obvious
TLDR: The Economist used AI to grade its past predictions and concluded it does okay, especially on less daring calls. Commenters were savage, joking that the headline answers itself with “yes” and mocking the whole exercise as paywalled AI-marked homework.
The Economist tried a bold bit of self-checking: it fed about 7,000 of its editorials into an artificial intelligence system to see how often its big predictions about the future actually came true. The magazine says the results were pretty reassuring overall. Safe, mainstream predictions did best, wild swings did worse, and a few middle-of-the-road calls — like saying Bitcoin might stick around back in 2013 — turned out spectacularly well. But the internet was not in a forgiving mood.
In the comments, readers basically treated this like a confession wrapped in a science project. The harshest verdict? "It’s always wrong when it really matters." Ouch. Another commenter went straight for the jugular with the meme-ready line that this was a rare case where a question headline is answered with “yes.” And then came the full internet triple-stack insult: one user sneered that this was Betteridge’s law wrapped in a paywall wrapped in an AI article pretending to be expert analysis — which is the kind of roast you can almost hear people forwarding to group chats.
Even the chart got dragged. One reader complained the graphic was so cluttered it was basically useless, turning a grand data exercise into a blurry cloud of dots. Another skipped the whole paywall drama by posting an archive link, which honestly says everything about the vibe. The real consensus? People aren’t just debating whether The Economist predicts badly — they’re questioning whether using AI to judge your own homework makes the whole thing look even sillier.
Key Points
- •The Economist used GPT-5.5 to analyze about 7,000 leader articles from this millennium and identified roughly 1,400 that contained falsifiable predictions.
- •The AI scored each prediction on contrarianism and accuracy, with multiple runs averaged.
- •The article says highly conventional forecasts tended to be more accurate, while the most contrarian forecasts were less accurate.
- •Examples of misses include forecasts on oil prices and oil demand, while cited successes include a 2007 call on Vladimir Putin’s continued rule and a 2013 call that Bitcoin had staying power.
- •The article says its forecasts performed best when they were neither very conventional nor extremely contrarian, claiming better-than-even odds of being correct in that middle range.