Pew Research Center's VP of Methods outlines two distinct data-quality threats to the polling industry: "silicon sampling" (substituting AI responses for real human ones) and large-scale AI-driven fraud in opt-in surveys, where bad actors can generate thousands of fake responses for financial reward. Internal and external studies find that AI-generated survey responses stereotype groups, underrepresent Republican viewpoints, and understate opinion disagreement, rendering them scientifically unreliable as proxies for public sentiment. For financial institutions that rely on consumer-sentiment surveys or commission opt-in polling for market research, these findings are a direct caution about data provenance and the validity of findings from non-probability-based survey vendors.
118 other tracked developments touching the same claims.