Can AI predict Polymarket questions?
Updated 22 October 2026
Yes, better than chance. Without ever seeing the market price, the average of 10 AI models picked the right side on 29 of 41 resolved questions (70.7%), and on 11 of 12 when the AIs were confident. The market itself is still the stronger forecaster overall (75.6%), which is why the interesting part is where the two disagree.
How the test works
- Every day: the most-traded open questions on Polymarket and Kalshi.
- 10 AI models each give a Yes probability, with a short web research brief and no price.
- The market price at that moment is stored next to the AI forecast.
- When the question resolves, both are scored.
The record so far
| AI (average of 10) | Market | |
|---|---|---|
| Right side, all questions | 29 of 41 (70.7%) | 75.6% |
| When the AI was confident (80%+) | 11 of 12 | – |
| Sports questions | 23 of 31 | 23 of 31 |
| Within 15 points of the market, blind | 31 of 41 | – |
| Best model vs market | DeepSeek V4 Pro beat the market on 21 of 41 | – |
How to use it
Treat the AIs as a second opinion, not a signal to trade. When ten independent models land close to the market price, the crowd and the machines agree. When they are far apart, read the AI reasoning and the market rules before drawing conclusions. See today's questions on AI vs Polymarket and Kalshi.
Questions
Is AI more accurate than Polymarket?
Not overall yet: on 41 resolved questions the market favourite was right 75.6% of the time against 70.7% for the AI average. Confident AI calls did better (11 of 12).
Is this financial advice?
No. TuringStats shows forecasts for information only and offers no trading.
Ask 10 AIs before your next match.
Free picks for every match, every model graded, and your own questions answered by the whole council.