AI / WORKING NOTE
I Asked Four AI Models to Predict the World Cup. Then Football Answered.
One question, four free AI tools, and a Spain-shaped consensus. A personal experiment in prediction, curiosity, and the emotions data cannot capture.
I asked AI. Then football answered.
Before the 2026 World Cup, I put the same question to four AI tools: Who will win? No paid plans. No coding. Just curiosity, a shared prediction framework, and ChatGPT, Gemini, Claude and DeepSeek.
This was a small personal experiment, not a forecasting benchmark. I wanted to see where the answers would differ—and where they would meet.
One question. Four answers.
My sketchnote captured the winning-team picks:
- ChatGPT: France
- Gemini: Spain
- Claude: Spain
- DeepSeek: Spain
Spain had three of the four votes. All four also named Mbappé as their Golden Boot favourite. Those were predictions recorded in the experiment, not guarantees or verified tournament outcomes.
Then the result arrived
Spain won the World Cup. The majority pick in my experiment matched the eventual champion, as confirmed by FIFA’s final tournament standings.
In my follow-up on LinkedIn, I wrote:
“AI found the patterns.
Football gave us the emotions.
Prediction is data. Celebration is human.”
That captured the feeling. Comparing answers was interesting; seeing the tournament unfold was something else entirely.
What this did—and didn’t—show
Three matching predictions made a memorable result. They did not establish that AI predicts football better than fans, analysts or a simple favourites list. There was no controlled comparison with those alternatives.
Models can share information and assumptions, so agreement is not necessarily independent evidence. One tournament is also too small a sample to judge reliability. The original posts do not document exact model versions, complete prompts or repeated trials, so this is a record of my experience rather than a reproducible test.
The experiment worth repeating
Next time, I’d save the exact prompt, model versions, dates, sources and probabilities before kick-off, then score every prediction—not just the winning-team pick. That would make the comparison more useful and less vulnerable to hindsight.
For me, the appeal remains simple: ask a better question, compare the reasoning, and stay curious about the result.
AI sees patterns. Football writes the story.
The original posts
More football and data: explore my 2022 FIFA Tableau dashboard.