* Prediction markets quadrupled annual trading volume to $63.5 billion in 2025.
* Three platforms (Kalshi, Polymarket, Opinion) now control over 95% of global volume.
* CertiK identifies oracle manipulation as the primary technical attack vector.
Prediction markets quadrupled their annual trading volume from $15.8 billion in 2024 to $63.5 billion in 2025, according to a new report from blockchain security firm CertiK. Activity stayed elevated after the US election cycle and extended into January 2026, with the week ending January 18 posting a record $6 billion in notional volume.
CertiK argues the next phase of growth runs into an integrity problem. The issue has less to do with smart contract exploits and more to do with the layers governing onboarding, volume authenticity, and the mechanisms determining who gets paid when markets resolve.
Three platforms now control more than 95% of global prediction market volume. Kalshi operates as a regulated venue in the US and positions itself as the compliance-first model. Polymarket captured the largest share of crypto-native and international participation. Opinion, the fast-growing entrant, scaled from effectively zero to roughly 30% market share in months using aggressive incentive programs.
A failure at any major venue no longer stays contained — it becomes a market-wide trust shock affecting liquidity pools, data feeds, and user balances. Brokers and mainstream distributors increasingly treat prediction probabilities as a new class of information product, which magnifies the consequences of platform-level failures.
Identity Failures and Web2.5 Tradeoffs Expose Centralization in a Decentralized Product
CertiK highlighted a December 2025 incident involving Magic.link, Polymarket’s third-party authentication provider. Accounts using Web2-style login methods — email or social authentication — were compromised, placing funds at risk. The on-chain settlement layer remained secure. The failure occurred at the identity layer, not the settlement layer.
The incident exposed the tradeoff inherent in “Web2.5” onboarding: smoother user experience in exchange for centralized failure points. Prediction markets can support fully collateralized on-chain settlement while retaining the same third-party risks plaguing conventional fintech — authentication, account recovery, and platform-level access controls. For an industry marketing itself on decentralization, the lesson lands uncomfortably.
CertiK also drew a line between trading volume as a proxy for adoption and probability outputs as a proxy for information. Incentive programs can inflate activity without improving forecasting quality.
The report cited research estimating artificial volume reached as high as 60% on some platforms during peak airdrop-farming periods. Wash trading remains widespread, which can mislead institutional users about liquidity depth and organic participation.
CertiK argues the more important question is whether probabilities remain useful even when the volume tape is noisy. In the report’s view, wash trading inflated volume metrics but has not yet compromised price accuracy. Probability outputs stayed reliable for forecasting. Platforms seeking to graduate to mainstream finance may position themselves as information utilities even if activity metrics are partly fabricated by incentives.
Beneath the headline numbers, CertiK described a structural rotation in how prediction market liquidity executes. Polygon retained dominance through the November election cycle, but BNB Chain volume surged beginning in late 2025, correlating with Opinion’s accelerated incentive rollout. By the week of January 19, BNB Chain activity flipped the historical hierarchy, capturing the plurality of weekly flows.
CertiK characterized oracle manipulation as the primary technical attack vector because market-resolution mechanisms directly control fund distribution. Ambiguous market definitions have already caused disputes across all major platforms throughout 2025, especially where political outcomes or contested official results create gray areas. As prediction markets scale, those tradeoffs become harder to ignore.
