Paper: Measuring User Responses to Age Verification Architectures: Evidence from a Deceptive Online Experiment
slides-agews-paper-measuring-user-responses-to-age-verification-architectures-evidence-from-a-deceptive-online-experiment-00

Yanzi Lin, Vivianna Lieu, Cheng Zhang, Weiqian Zhang, Lorrie Faith Cranor, Sarah Scheffler

Following the U.S. Supreme Court’s decision in Free Speech Coalition v. Paxton (2025), which established that age verification systems must be “adequately tailored,” understanding user behavior has become legally relevant for system design. This preliminary study empirically examines how different age verification methods affect user behavior through a deceptive online experiment framed as usability testing for a mock gambling website. Participants (n=99 U.S. residents) were randomly assigned to six verification conditions, including simple checkbox self-declaration, government-issued ID upload, and AI-based facial age estimation. Results show stark differences in user responses: checkbox verification achieved 95.2% completion rates, while government ID methods drove up to 60.5% of users to return their study without finishing. We also tested the effects of privacy disclosures on completion rates. These had mixed effects, with detailed data handling information both increasing completion rates and polarizing user comfort levels. In a survey accompanying the empirical study, participants expressed significant privacy concerns about document- based methods, citing fears of identity theft and data misuse. These findings provide empirical evidence that can be applied to the U.S. Constitutional requirement for “adequate tailoring” of age verification systems, as well as policy analysis and technical design of age verification more broadly. We outline plans for expanded research using R-rated movie content to examine these effects at larger scale.