TRM Labs Analysis Shows Most x402 Payments Do Not Come From AI Agents
New blockchain analysis reveals that only a very small fraction of payments passing through x402, a machine-to-machine commerce protocol, consists of autonomous artificial intelligence agents.
A new study conducted by blockchain intelligence firm TRM Labs showed that a very small proportion of transactions on the x402 protocol, designed for machine-to-machine commerce, belong to autonomous AI agents.
Structure of the x402 Protocol
Launched by Coinbase in 2025, the x402 protocol transforms the HTTP 402 payment status code into an internet-native payment mechanism.
Although the protocol eliminates traditional payment pages, it does not require artificial intelligence to operate, and ordinary software scripts can execute the same process.
Analysis Conducted by TRM Labs
TRM Labs examined approximately $52.7 million in value across 198.9 million x402 settlement transactions on the Base, Solana, and Polygon blockchains starting from May 2025.
After filtering out movements that do not represent genuine commerce, it was estimated that only between 0.6 percent and 7.5 percent of the remaining payment value originated from agents.
Detection of Agent Behaviors
Since agent identity is not directly visible in on-chain data, TRM based its criteria on factors such as discovering different products and making payments to multiple vendors.
While addresses continuously paying the exact same amount to a single vendor were evaluated as traditional automation, this situation may also lead to some real agents being overlooked.
Stablecoin and Compliance Issues
The examinations proved that almost all x402 transactions were carried out using USDC, demonstrating that the system is stablecoin-focused.
Agent payments combining very low dollar values with high transaction volumes pose new challenges for existing compliance and regulatory mechanisms designed around human payment habits.