AI Polling Stress Test

Courtney Kennedy, vice president of methods and innovation, answers some common questions about the current polling landscape in the U.S.

1. Frame 1Polling's public promise faces a risk as AI answers and fake respondents test how surveys find real people.
2. Frame 2Silicon sampling asks models to imitate voters instead of interviewing people, raising ethical and scientific concerns.
3. Frame 3Probability polling limits fraud because the path starts offline, using U.S. home addresses and mailed invitations.
4. Frame 4The opt-in path lets people self-enroll through reward ads, creating a fake-identity fraud bottleneck before data checks.
5. Frame 5Worked example: five bot accounts take 200 surveys daily at $1 each, grossing $30,000 monthly.
6. Frame 6The review chain checks recruitment, weighting, web-or-phone access, and data quality before trusting results.