MW We asked AI to predict the 2026 World Cup winner. It picked something that has never happened.
By Weston Blasi
Some 2026 World Cup predictions from AI models are going against tradition to pick the underdogs. Here's why.
Portugal forward Cristiano Ronaldo is set to participate in his sixth FIFA World Cup - and a few AI models really like his team's chances.
Can AI predict the winner of the 2026 FIFA World Cup?
The World Cup, which began Thursday in Mexico City, could top $60 billion in total wagers, up from the more than $35 billion bet during the 2022 tournament in Qatar. This jump is likely due to the exploding popularity of sports betting since 2022, as well as the expanded soccer tournament being hosted across the U.S., Mexico and Canada for the next 39 days, which now features 40 more matches than it did four years ago.
But unlike many other major sporting tournaments, like March Madness, the World Cup is historically much easier to predict. That's because only eight nations have ever won the tournament in its 96-year history: Argentina, Brazil, England, France, Germany, Italy, Spain and Uruguay. Still, the millions of fans trying to guess which team will win are likely looking for ways to narrow down their choices even more.
If you want to have some skin in the game while watching the world's biggest sporting event, but you didn't catch much soccer this year, then you might be looking for some projections on which team is favored to win before placing your bets. So we asked a few popular AI chatbots to predict which country will lift the 18-karat gold trophy awarded to the World Cup winners.
Yet when prompted to predict the winners of each individual matchup while being "as accurate as possible" for the 2026 World Cup, Google's $(GOOGL)$ $(GOOG)$ Gemini (Gemini 3.5 Flash) picked a tournament outcome that has never happened before.
Gemini picked the Netherlands as its predicted champion over reigning champion Argentina, led by star player Lionel Messi. Were this to happen, it would be the first time the Netherlands has ever won the World Cup, and the first new winner since 2010.
What's more, the projection had several other teams that have never before won the tournament advancing to the quarterfinals, including Portugal and Belgium. All of those nations have long betting odds to win the 2026 tournament, depending on the sportsbook.
Why was Gemini so bullish about these underdogs? The Google model provided detailed breakdowns of the Netherlands' stingy defense, strong midfield and easy schedule as part of its reasoning for the upset.
So what do professional human prognosticators have to say about this unexpected outcome?
"The favorites are the favorites for a reason," Anika Howard, CEO of online betting and technology company Wondr Nation, told MarketWatch about the historical predictability of the World Cup. "But what's great about the human experience is: Anything can happen."
Alan Levy, co-founder and CEO of 4C Predictions, a marketplace for sports, stocks and politics, was a bit more skeptical. "The matches are unpredictable," he noted. "The champion usually isn't."
Most of gambling public is still backing the traditional proven winners to prevail, unlike the projection from Google's Gemini. On BetMGM $(MGM)$, the most wagered-on teams to win the tournament are former winners Spain, with 25% of the money wagered, and France, with 22% of total money wagered.
FanDuel parent company Flutter $(FLUT)$ estimated it will see "a hundred thousand bets a minute coming through the platform" during matches. Spain and France are the two biggest favorites to win the 2026 World Cup on FanDuel, too.
But Levy said his AI models are also going for teams who have never won the World Cup before.
"The teams attracting the most interest in our models are the usual suspects," Levy said. "But the most interesting finding isn't at the top of the table: It's Portugal."
He said that his AI models project a strong performance for Cristiano Ronaldo's Portugal at the World Cup, and a chance to win it all. A World Cup win for Portugal would be the first in the nation's history.
"Portugal has something most nations don't: world-class talent, depth, experience, and enough quality across the entire squad to survive a month-long tournament," Levy said. "Many fans still think of Portugal as Cristiano Ronaldo's team. Our models think that's outdated. This may be the deepest and most complete Portugal squad we've seen."
In addition to Portugal, another nation targeted by Levy's AI model is Morocco - representing a continent, Africa, that has never seen one of its nations hoist the trophy.
"If our AI agents were forced to pick one country that is being underestimated, it would be Morocco. Not because Morocco has the best players, not because Morocco has the deepest squad, but because Morocco has already shown it can survive against the world's elite on the biggest stage," Levy added. "Most Cinderella stories end when they meet a giant. Morocco has already proven it can beat giants."
Of course, every AI model is different, and different large language models can provide users with distinct results. When given the same prompt, OpenAI's ChatGPT (GPT-5.5) produced a bracket with nearly every favorite advancing, ending with Argentina beating England in the final. Anthropic's Claude (Sonnet 4.6) similarly picked a quarterfinals featuring big nations including Spain, England, France and Brazil, with France as the predicted winner.
Upsets in the World Cup do happen every year in individual matches, but surprise tournament winners are far more rare. In the 2022 World Cup in Qatar, for example, Argentina lost to big underdog Saudi Arabia in the group stage - but it still went on to win the World Cup for the third time.
-Weston Blasi
This content was created by MarketWatch, which is operated by Dow Jones & Co. MarketWatch is published independently from Dow Jones Newswires and The Wall Street Journal.
(END) Dow Jones Newswires
June 11, 2026 17:37 ET (21:37 GMT)
Copyright (c) 2026 Dow Jones & Company, Inc.
Comments