Are Billions in Prediction Market Bets Actually Producing Useful Information
Prediction markets have arrived at scale — Polymarket and Kalshi now process billions in monthly volume. But a rigorous data analysis from Asterisk Magazine challenges the “truth machine” narrative.
The hard numbers. Over 80% of Polymarket’s volume is concentrated on sports betting, crypto prices, and election gambling. Even among markets categorized as “useful,” volume has plateaued since late 2024, and median trading volume is actually declining. After filtering nearly 200,000 markets, only about 7,000 qualified as potentially useful for decision-making.
Five categories, one winner. The analysis breaks prediction market value into five areas: risk monitoring, news interpretation, policy outcomes, accountability, and novel information. Risk monitoring — especially for geopolitical conflicts — genuinely works and gets cited by mainstream media. But the other four categories show limited evidence of producing actionable information.
The AI challenge. The article makes a compelling argument that AI chatbots may outperform prediction markets as forecasting tools. ChatGPT and Claude already serve as primary information sources for many professionals, offering not just probabilities but narratives, context, and follow-up questions — capabilities prediction markets inherently lack. The author’s company, FutureSearch, deployed AI forecasters in prediction markets as early as 2024.
Volume ≠ accuracy. Higher-volume markets lasting over 90 days are more accurate, but short-term markets show no statistically significant correlation between volume and accuracy. Useful market accuracy actually declined after early 2025.
For anyone tracking the evolution of information markets in the AI era, this article provides essential data-driven perspective — and a sobering reality check on the gap between prediction market hype and actual utility.