_12-45-2026_06-45.jpg)
In 2010, a sea creature named Paul ate his way into football history. He picked winning teams by choosing food boxes marked with national flags. Eight matches. Eight correct calls. Zero training data.
Sixteen years later, the oracles have changed. They run on graphics cards now, not saltwater tanks.
The 2026 FIFA World Cup is the first tournament of the generative AI era. ChatGPT went public just days after Argentina lifted the trophy in Qatar in 2022. Back then, almost nobody outside Silicon Valley had heard of it.
This time, everyone has access. And everyone is asking the same thing — who wins?
Every Major AI Has Picked a Side
The predictions are in. They do not all agree.
ChatGPT and Claude both tip Spain. They point to Spain's tactical discipline, squad depth, and strong recent form across major competitions.
Le Chat — the chatbot from France's homegrown AI company Mistral — backs France. It credits Les Bleus with enough firepower and tournament experience to go all the way.
DeepSeek and Qwen, the two leading Chinese AI models, both favour Argentina. They highlight Argentina's current world number one ranking and the clinical finishing of striker Lautaro Martínez.
Tech publication Tom's Guide ran its own test. It asked Google Gemini, ChatGPT, and Perplexity AI the same question. All three returned Spain as champion. All three named France as runners-up.
News outlet Decrypt found the same western consensus. But it also confirmed what Chinese models are saying — DeepSeek and Qwen both break from the crowd and go with Argentina.
Bank of America joined the experiment too. Their analysts put the question to Microsoft Copilot. The chatbot split its answer evenly between Spain and France — matching roughly 40 percent of surveyed fans who also backed France.
When investment banks start using AI chatbots to call football results, this has clearly moved beyond novelty.
Why Do Different AIs Pick Different Winners?
This divergence is not random. It tells you something important about how these systems work.
Every AI model learns from a different diet of text. Western models consume enormous amounts of English-language sports journalism. That coverage heavily favours Spain and France as footballing powers.
Chinese models train on a different media landscape. In that world, Argentina's 2022 World Cup triumph was enormous news. The data reflects that.
The result: an AI's World Cup prediction reveals its training sources as clearly as its football knowledge. It is not a bug in the system. It is the system working exactly as designed.
Germany's Researchers Are Keeping Score
While fans debate predictions, researchers at Germany's Ludwig Maximilian University are building something more rigorous. Their project is called LLM SoccerArena, and it runs a live public leaderboard throughout the tournament.
Every AI model makes a game-by-game prediction. Every result checks that prediction against reality. The accuracy data goes public immediately.
LMU researcher Stefan Feuerriegel explained why this matters. He said the core question is whether language models can reliably support real decisions. He argued that any meaningful benchmark must test how AI handles dynamic information and genuine uncertainty — not just tidy, fixed tasks.
The LMU team tests two distinct prediction modes:
Mode 1 — Training data only: The AI answers from what it already knows. No live updates. No real-time context.
Mode 2 — Web-augmented: Before predicting, the AI searches for current injury reports, squad announcements, and live betting odds. It incorporates that information before giving a verdict.
Mode 2 consistently outperforms Mode 1. This matters because football is a sport where a pulled hamstring on a Tuesday can change everything by Saturday.
Paul vs the Machines: A Fair Comparison?
Paul the Octopus worked at a German aquarium called Sea Life Oberhausen. During the 2010 World Cup in South Africa, he predicted Germany's matches by choosing which of two flag-marked food containers to eat from first.
He got every single one right. Seven Germany matches. One final. Eight for eight.
He also called Spain as tournament winners before the final — correctly.
Paul never gave reasons. He had no injury data, no squad sheets, no tactical analysis. He had mussels.
AI systems work entirely differently. They generate probabilistic forecasts with supporting logic. You can ask them why they picked Spain and receive a structured argument. You cannot do that with an octopus.
But there is a strange humility in acknowledging this: Paul's hit rate was perfect. The AI leaderboard at LMU SoccerArena is still being written.
AI Is Doing Far More Than Making Predictions
Forecasting winners is only the most visible way AI shapes this World Cup.
Researchers from Australia's University of the Sunshine Coast documented the broader picture in an article for The Conversation. AI now operates quietly across almost every layer of the tournament.
Coaching staff use it to analyse opponent tactics and build match-day game plans. Medical teams use it to track player workloads and reduce injury risk. Referee review systems use AI to flag incidents for closer examination. Even tournament security teams use it to identify fraudulent ticket activity before fans reach the gates.
The researchers put it plainly: no AI will score a goal or call a substitution this summer. But every team that reaches the final will have relied on AI tools somewhere along the way.