"Researchers report that up to 45% of survey responses are potentially AI-generated. The use of LLMs in social science research threatens the validity of behavioral and social science data. Scientists are implementing 'honeypots' and other detection methods to combat AI-generated survey pollution."