Little research about the analysis of web applications’s response has been done . Its level of automation and intelligence is very low. And web penetration still requires human intervention which directly limit its efficiency . To solve this problem,key words analysis and negative response extraction have been studied, and with the help of the two technology a algorithm named self-learnning response analysis is proposed. At first the algorithm use key words to ananlyse the response. If it fails, heuristic analysis technology is used to work it out. The key words extracted from the response will be stored in the database which is used to keep key words. Experimental results show that the algorithm can analyse the response quickly and correctly. And it can analyse the response that key words analysis can not. At the same time, it is more efficient than negative response extraction.