How Four AIs Predicted the 2025 F1 Champion — And Why the Margin Matters
- Chris Howell
- 2 hours ago
- 4 min read
When Mercia AI asked four large language models — ChatGPT, Gemini, Grok and Perplexity — to run Deep Research to predict the 2025 Formula 1 World Champion before a single race had been run, the aim was not to prove that AI can “see the future”. The experiment was never intended as a party trick, a headline grab, or a claim of predictive certainty....well ok, yes, we had fun with it — but underneath, this was always about pattern-spotting and decision support.
The real question was simpler, and far more practical:
Can AI spot long‑term patterns earlier than humans — and can those patterns hold up under real‑world pressure?
Months later, after one of the closest and most dramatic season finales in recent Formula 1 history, the answer arrived in the narrowest possible way.
Lando Norris is World Champion — by just two points.
Crucially, all four AIs — working independently — had predicted him before the season began.
This wasn’t a comfortable win. It wasn’t a dominant campaign. It wasn’t even decided until the final laps of Abu Dhabi, with strategy calls, reliability scares, and championship‑defining moments unfolding right to the chequered flag. Yet despite all that noise, the models had converged on the same outcome months earlier.
So what does a two‑point margin really tell us about AI, prediction, and decision‑making in the real world?
1. The prediction wasn’t certain — it was directional
AI didn’t say Lando will win because we know the future. It said Lando is the most likely champion given the available information at the time.
That distinction matters.
The models were not running statistical simulations, Monte Carlo forecasts, or race‑by‑race scenario modelling. They weren’t ingesting live telemetry, tyre data, or hidden performance metrics. Instead, they reasoned over context, history, and long‑range patterns when forming their answers.
In their explanations, the AIs’ predictions often referenced factors such as team development, driver form, reliability trends, and historical season patterns when explaining why Norris emerged as their preferred choice.
That context commonly included themes such as:
Team stability and development trajectory over recent seasons
Driver consistency and performance progression
Reliability patterns and risk exposure
Relative momentum compared to close championship rivals
The likelihood of sustained competitiveness across a full season
But it did not include:
Lap‑by‑lap race data
In‑race incidents or safety cars
Mechanical failures
Track‑specific anomalies
Late‑season psychological pressure
McLaren dropping vital points in Las Vegas (plank issues) and Qatar (a strategy call backfiring after a Lap‑7 safety car).
In other words, the models formed a probabilistic, long‑range forecast, not a promise.
That’s precisely why the outcome matters.
In the end, the chaos didn’t invalidate the prediction. It validated it.
A narrow win doesn’t make the forecast lucky. It makes it honest.
2. Why a two‑point title tells us more than a dominant one
If Norris had wrapped up the championship with races to spare, the analysis would be easier — and far less interesting.
A dominant season can mask uncertainty. It smooths over complexity and makes hindsight feel obvious. A two‑point margin does the opposite.
It tells us that:
The outcome was genuinely uncertain until the very end
Small variables could easily have flipped the result
Narrative alone was not enough to identify the champion
Underlying trends still mattered despite the chaos
A comfortable win can be predicted by many models. A razor‑thin win suggests the reasoning was doing real work, along with McLaren's pit-stop team.
This is where AI shows its value: not by guaranteeing outcomes, but by identifying who is most likely to succeed when margins are tight and uncertainty is high.
3. Forecasting is about reducing uncertainty, not eliminating it
Every Formula 1 strategist — and every business leader — faces the same fundamental challenge:
You have to make decisions using incomplete information.
AI forecasting doesn’t remove uncertainty. It helps you understand and navigate it.
Used well, AI can:
Surface patterns humans overlook
Highlight the most likely scenarios
Expose hidden assumptions
Clarify risk rather than ignore it
Improve decision quality over time
AI doesn’t replace judgement. It supports it.
That’s as true in the boardroom as it is on the pit wall.
4. What this prediction teaches us about real‑world AI use
The real success of this F1 experiment wasn’t simply that AI got the champion right.
It was that AI applied structure and reasoning to a messy, unpredictable system — and still produced a useful, resilient forecast under extreme conditions.
That same approach applies far beyond sport. Organisations face similar challenges when dealing with:
Demand forecasting
Resource planning
Risk assessment
Portfolio decisions
Long‑term strategic choices
AI is most valuable when problems are complex, noisy, and full of competing variables.
That’s where humans struggle. That’s where AI helps most.
5. Beyond sport: what organisations can take from this
You don’t need AI to be perfect.
You need it to be:
Consistent
Explainable
Directionally accurate
Grounded in context
If AI can help surface the right outcome in a Formula 1 season decided by two points, it can also help organisations make better calls when the stakes are high and the margins are small.
A note on Deep Research capability
Since the original prediction blog was published, Deep Research capabilities across several large language models have continued to evolve. Improvements in how models gather, synthesise, and structure information have made their reasoning more transparent, more explicit, and easier to interrogate — particularly when comparing outputs across multiple models.
These advances don’t turn AI into a crystal ball, and they don’t eliminate uncertainty. What they do provide is clearer long‑range reasoning, better‑structured explanations, and more useful insight when exploring complex, uncertain scenarios where traditional certainty simply isn’t available.
Applying these ideas to your own data
If you’d like to explore how AI can help you spot patterns, reduce uncertainty, and make more informed decisions using your own data, Mercia AI offers practical support through:
Starter Data Insight — turning existing data into clear trends and actions
AI Readiness Consultation — identifying where AI can genuinely help your organisation
These services focus on real‑world clarity, not hype — the same principle that guided this prediction experiment.
