Prediction markets were built to aggregate human forecasts, but an increasing share of trading opportunities is being captured by automated systems—now often powered by AI agents. Short-lived mispricings, such as markets whose probabilities don’t sum to 100% or delays in updating after events, open tiny windows for arbitrage that machines can scan for and exploit far faster than any person.
Rodrigo Coelho, CEO of Edge & Node, notes that bots already monitor hundreds of markets per second and that AI-driven agents are taking on more of that workload. Capturing these fleeting opportunities requires watching thousands of markets and executing trades almost instantly, which is why automated systems dominate in this space. Prediction markets are therefore a natural target for AI designed to detect and act on ephemeral pricing gaps.
Latency is often the deciding factor in prediction-market arbitrage. If a market takes even a few seconds to reflect a new piece of information, execution bots can register the correct outcome and place bets within that window. An academic study of Polymarket found frequent inconsistencies both within single markets (probabilities not totaling 100%) and across related markets, estimating roughly $40 million extracted from those inefficiencies. Platforms have responded with measures such as taker fees or delayed finalization of outcomes, which can blunt but not eliminate some strategies.
AI agents also raise manipulation concerns. Thinly traded markets are sensitive to large wagers, and more capable agents could scale or repeat market-moving actions. Coelho pointed to past examples where a single large bet affected high-profile political markets. Polymarket’s open interest peaked around the 2024 U.S. election and remains elevated, with politics, sports and crypto among the most active topics.
Pranav Maheshwari of Edge & Node warns that as agents become more capable, stronger guardrails will be required. Until now, many agents have had only medium capability and broad permissions, and even those have begun acting autonomously. Future, higher-capability agents will need tighter restrictions on what they can access and do.
Trading automation is also evolving. Execution bots started with simple, rule-based scripts; now systems increasingly combine AI for idea generation, signal detection and risk assessment with automated execution. Archie Chaudhury, CEO of LayerLens, says most retail traders still rely on chatbots like ChatGPT or Gemini for research, while advanced users employ coding agents (for example Claude Code) or autonomous execution tools (such as OpenClaw) to design and run full trading strategies.
As AI literacy spreads, some strategies once reserved for institutions may become accessible to retail traders, though competition will intensify. Major firms already use private AI systems to trade, and current large-language-model architectures make it easier to interpret structured financial data, lowering the technical barrier for building automated systems that previously required specialized quantitative skills.
These shifts mirror the broader crypto markets, where arbitrage and edge increasingly depend on automation rather than human judgment. Speed and integrated AI-driven execution are becoming the primary competitive advantages. That dynamic suggests that market participants who adopt advanced automation will likely outperform slower, manual actors—and it underscores the need for platform safeguards and regulatory attention as AI agents play a larger role in prediction-market activity.