New research reveals that artificial intelligence can now corrupt public opinion polls at scale, passing all quality checks and mimicking human responses without detection. This poses a severe threat to election predictions, scientific research, and the integrity of public discourse.
The Vulnerability of Polling
A study from Dartmouth, published in the Proceedings of the National Academy of Sciences, demonstrates how easily AI can manipulate survey results. Researchers created a simple AI tool—an “autonomous synthetic respondent”—that operates from a 500-word prompt and successfully evaded detection in 99.8% of attention checks designed to identify automated responses. The tool made zero errors on logic puzzles and tailored responses to randomly assigned demographics, making the data appear entirely legitimate.
How Easily Results Can Be Flipped
The implications are stark: adding as few as 10 to 52 fake AI responses to major national polls before the 2024 election could have flipped the predicted outcome. When programmed to favor either Democrats or Republicans, presidential approval ratings swung from 34% to either 98% or 0%, and generic ballot support went from 38% Republican to either 97% or 1%.
The Financial Incentive and Lack of Detection
The financial incentive to use AI is significant. Human respondents typically earn $1.50 for completing a survey, while AI bots can complete the same task for approximately five cents. A recent study found that 34% of respondents had already used AI to answer open-ended survey questions. Critically, every current AI detection method failed to identify the AI tool used in the Dartmouth study.
Beyond Elections: A Crisis for Research
The problem extends far beyond election polling. Surveys are fundamental to scientific research across disciplines, including psychology, economics, and public health. Thousands of peer-reviewed studies rely on survey data to inform research and shape policy. If survey data is poisoned by AI, the entire knowledge ecosystem is at risk.
The Need for Transparency and Verification
The study argues for transparency from companies that conduct surveys, requiring them to prove their participants are real people. New approaches to measuring public opinion are needed—approaches designed for an AI world. The technology exists to verify human participation; the critical missing piece is the will to implement it. If action is not taken now, the integrity of polling and democratic accountability will be further eroded
