According to CoinMeta, Wu said that Circle has published a research paper proposing a reputation evaluation system for the AI and agent service markets, aimed at solving issues such as fake ratings, volume manipulation, and “witch attacks.” This system is based on X402 on-chain payment records and generates personalized reputation rankings by analyzing the actual financial transactions between buyers and service providers, making it necessary for attackers to pay real funds to independent service providers in order to improve the rankings of their fake services. The system was tested using 66.1 million X402 payments covering 16 blockchains with a total value of approximately 6.1 million US dollars. The results show that the accuracy of the recommendations from this system is close to that of traditional collaborative filtering algorithms, but under certain attack models, the cost of manipulating rankings is about 23 times that of the latter methods. This system does not require any modification to the existing X402 or ERC standards, nor does it need to introduce new tokens.