Reply to Westwood: Questioning the empirical evidence that AI survey contamination is real and substantial

Westwood [2025], followed closely by Van der Stigchel et al. [2026] and Westwood and Frederick [2026], argues that “AI contamination” poses a “potential existential threat of large language models to online survey research.” Although AI (frequently LLMs) poses potential challenges for survey research, the articles overstate their case, conflating distinct risks and advancing claims of field-level vulnerability without (1) clear definitions, (2) appropriate benchmarks, or (3) reproducible demonstrations of real-world impact.