8 AI Models Predicted the 2026 World Cup. Here's How Accurate They Were.

Published July 20, 2026 by Aldo Giammusso

Back in June we asked eight AI models to call the World Cup. They near-unanimously crowned France. Spain just lifted the trophy, so we graded every pick against the real results.

In June, before a ball was kicked, we ran an experiment: we asked eight AI models, ChatGPT, Claude, Copilot, Gemini, Grok, DeepSeek, Perplexity, and Meta AI, to predict the 2026 World Cup. We ran it twice, once cold and once after feeding every model the same dataset, and we wrote up all sixteen forecasts in our original World Cup prediction post.

The headline back then was how much the models agreed. France was the runaway pick. Seven of eight chose France in the cold round, and seven of eight chose France again after seeing the data. It was the strongest consensus in the entire experiment.

Now the tournament is finished. Spain beat Argentina 1-0 after extra time on July 19 to win their second World Cup, France finished fourth, and we can grade every single pick against what actually happened. The result is more interesting than a simple right or wrong. The models were confidently, near-unanimously wrong about the one thing everyone remembers, and quietly sharp about almost everything else.

The One-Line Verdict

The AI consensus got the trophy wrong and the structure right. Fourteen of the sixteen forecasts crowned France. France finished fourth. Only one forecast in sixteen, Meta AI in the data-fed round, named the actual champion. Yet the same models correctly pinpointed the four semifinalists, nailed the Golden Boot, and flagged the tournament's biggest flop before it happened.

The Report Card

Here is how the AI consensus stacked up against the real 2026 World Cup, category by category.

Where the AI Models Nailed It

The Golden Boot was a clean hit. The single most confident individual call in the whole experiment was Kylian Mbappé to finish top scorer, and he did, with 10 goals. He also passed Lionel Messi to become the all-time leading World Cup scorer on 22 career goals. Fourteen of the sixteen forecasts had him down for it. When the models agreed on something rooted in a clear, measurable strength, a great striker on a strong team, they were right.

They pinpointed the final four. The real semifinalists were Spain, France, Argentina, and England. Those four names showed up in the models' top brackets constantly. Claude named all four semifinalists in both rounds. Gemini and Perplexity named all four in the cold round. The models could not tell you which of the four would win, but they could tell you it would be those four, and it was.

They called the biggest flop. In the cold round, Claude, Grok, and Meta AI all flagged Germany as the team most likely to exit early, and Meta AI even pinned it to the Round of 32. Germany went out in the Round of 32, losing a penalty shootout to Paraguay. In the data-fed round, four models tagged Germany as the most overrated team in the field. Claude also flagged Brazil as overrated in the cold round, and Brazil crashed out in the Round of 16 to Erling Haaland's Norway. Norway, in turn, was named the tournament's biggest surprise by three models in the data-fed round, and they were right about that too.

France was not a crazy pick. It is worth being fair to the models here. France reached the semifinals and were one of the last four standing. The AIs did not put their money on a weak team. They put it on a genuinely strong one that went deep. They just missed the finish.

Where the AI Models Missed

The champion. This is the big one. France appeared as champion in fourteen of sixteen forecasts and finished fourth, losing the semifinal to Spain and then losing the third-place game to England 6-4. The one call everyone was watching, and the consensus blew it.

Best Young Player. Eleven of sixteen forecasts named Lamine Yamal. The award went to Spain's 19-year-old defender Pau Cubarsí. Not one model named him.

Golden Ball. This was the category the models were least sure about in the first place, naming five different players. It went to Spain's Rodri, and here is the kicker: not one of the sixteen forecasts named Rodri at all. The category they were least confident about, they still got completely wrong.

The scoreline. Almost every model predicted a 2-1 final. Spain won 1-0. In fairness, several models sensed a tight, dramatic final and predicted extra time, and the final did go to extra time. They read the tension right and the numbers wrong.

The Lonely Outlier That Got It Right

Every experiment like this has one detail that sticks. Here it is: the single forecast that named Spain as champion was Meta AI, in the data-fed round. It was the model the consensus disagreed with. Seven other models looked at the same data and stayed with France. One looked at it and switched to Spain, and that one was right.

That is the most useful thing in this entire experiment. The loudest agreement in the room was wrong, and the quiet outlier was correct. If you had simply gone with the crowd, you would have picked France. If you had paid attention to why one model broke from the pack, you might have caught the actual winner.

What Actually Happened at the 2026 World Cup

For the record, and because these are the facts every one of those predictions is now measured against:

• Champion: Spain, beating Argentina 1-0 after extra time. Ferran Torres scored the only goal in the 106th minute. It was Spain's second world title.
• Final four: Spain, France, Argentina, and England. For the first time, the four top-ranked teams in the world all reached the semifinals.
• Third place: England beat France 6-4 in a ten-goal thriller, the most goals in a World Cup match since 1982, with a Bukayo Saka hat trick.
• Golden Boot: Kylian Mbappé, 10 goals, his second in a row. Golden Ball: Rodri. Best Young Player: Pau Cubarsí. Golden Glove: Unai Simón. Three of those four awards went to Spain.
• Biggest upset: for the first time in World Cup history, neither Brazil nor Germany reached the quarterfinals. Brazil went out in the Round of 16, Germany in the Round of 32.
• The Cinderella story: Cabo Verde, a nation of about half a million people, reached the knockout stage, the smallest country ever to do it.
• The hosts: the United States, Canada, and Mexico were all knocked out in the Round of 16.

What This Actually Tells You About AI

This was a fun experiment, but the pattern in the results is the real payoff, and it applies to how any business should use AI.

AI is excellent at aggregate pattern-matching. Ask it which teams are strong, how far they are likely to go, or who is due for a disappointment, and it does a genuinely good job, because those questions are about weight of evidence. That is exactly why the models nailed the semifinalists, the Golden Boot, and the German collapse.

AI is weak at single discrete outcomes and subjective calls. Which one team out of the last four lifts the trophy? Which player a committee decides was the best? Those are not pattern questions, they are coin-flip and judgment questions, and the models were confidently wrong on every one of them.

So the lesson for a business is not "AI is smart" or "AI is dumb." It is this: use AI for the trend, the shortlist, and the pressure test, and keep a human for the final call. That is precisely how we use AI at AldoMedia. It drafts, it researches, it spots patterns across a mountain of data faster than any person could, and then a human decides what actually ships. The same tool that picked the wrong champion picked the right semifinalists, and knowing which is which is the entire job.

It is also worth remembering that the AI everyone is now searching with, the same models in this experiment, is deciding which businesses to recommend to real customers. Getting your site understood by those tools is a new and very real part of being found online.

Frequently Asked Questions

Did any AI model predict Spain to win the 2026 World Cup?

Yes, but only one forecast out of sixteen. In the data-fed round, Meta AI was the only model to name Spain as champion. Fourteen of the sixteen forecasts picked France, who finished fourth, and one picked Brazil, who went out in the Round of 16.

Who actually won the 2026 World Cup?

Spain won the 2026 World Cup, beating Argentina 1-0 after extra time on July 19, 2026, at MetLife Stadium in New Jersey. Ferran Torres scored the only goal in the 106th minute. It was Spain's second world title and their first since 2010.

Which World Cup prediction did the AI models get right?

The Golden Boot. Fourteen of the sixteen forecasts named Kylian Mbappé as the tournament's top scorer, and he won it with 10 goals. The models also collectively identified the four semifinalists and correctly flagged Germany as an overrated team that would exit early.

Did the AI models predict France to win the World Cup?

Yes, overwhelmingly. Seven of the eight models picked France in both the cold round and the data-fed round. France reached the semifinals but lost 2-0 to Spain, then lost the third-place match 6-4 to England and finished fourth.

What did the AI models get most wrong?

The champion and two of the individual awards. France appeared as champion in fourteen of sixteen forecasts and finished fourth. Eleven of sixteen forecasts named Lamine Yamal for Best Young Player, which went to Spain's Pau Cubarsí, and not one of the sixteen named Rodri, the actual Golden Ball winner.

Can AI reliably predict sports results?

No. AI models pattern-match on team strength, history, and the data you give them. They are good at spotting which teams are strong and how far they are likely to go, which is why they pinpointed the semifinalists and the Golden Boot. They cannot predict a single discrete outcome like which one team lifts the trophy, and they were confidently wrong about that.