AI's Threat to Polling Exposed by Withdrawn Church Data
If you had been following recent news about church attendance in Britain, you might have believed the nation was experiencing a Christian revival. Stories of swelling congregations, particularly among young people returning to the flock, seemed to gain credibility from a 2024 report by the Bible Society. This report, based on data collected through a YouGov survey, claimed that church attendance was increasing in England and Wales, driving headlines and establishing a compelling narrative.
However, there was a critical flaw: the survey has since been withdrawn due to fraudulent data. Academics and experts now warn that this incident serves as a cautionary tale, not about a religious renaissance, but about the dangers posed by artificial intelligence to polling integrity. They argue that online opt-in surveys are increasingly plagued by bogus responses, as participants often use AI tools to complete questionnaires quickly, sometimes for financial gain.
The Growing Problem of Survey Fraud
David Voas, a quantitative social scientist and emeritus professor at University College London, emphasized the difficulty in correcting such misinformation once it spreads. "This sort of information, misinformation, is just very difficult to correct once it starts spreading," he said. "And the amount of effort required to correct it is an order of magnitude higher than the effort needed to disseminate it in the first place." Voas noted that this issue undermines confidence in polls and is not isolated to YouGov, but is a growing concern globally, as participants can generate significant revenue by working at scale.
According to a 2026 report, YouGov employs a random sample method to prevent participants from selecting surveys, but the rise of AI has exacerbated vulnerabilities. Sean Westwood, an associate professor at Dartmouth College, stated that the fundamental assumption of survey research—that respondents are real people giving coherent answers—is now broken. "The tools to do all of this are cheap, accessible and available right now," he warned, highlighting AI's potential to bias results invisibly.
Challenges in Detection and Youth Influence
Westwood explained that AI models can be weaponized to manipulate responses, such as in political polls, while maintaining demographic profiles to evade detection. "Even without explicit instructions to cheat, the agent can figure out what a researcher is trying to test and produce data that confirms the hypothesis," he said. However, the exact scope of AI use in surveys remains unclear, and rapid technological advancements make it difficult to combat, as fixes become obsolete quickly.
The Bible Society report specifically claimed that church attendance was rising fastest among young people. Courtney Kennedy, vice-president of methods and innovation at Pew Research Center, noted that opt-in estimates for those under 30 often contain high error rates and are more likely to originate from click farms. "Bogus respondents tend to respond in the affirmative, no matter what is asked," she said, describing a positivity bias that inflates estimates. Voas added that the report's failure to critically compare findings with other church research further compounded the problem.
Industry Responses and Future Implications
A YouGov spokesperson addressed the issue, stating that the company uses various methods to detect fraud, including identity checks, device fingerprinting, and real-time threat scoring. "The rise of organised survey farms, bots, and now AI-assisted responses makes detection a vital, continuous and constantly evolving discipline," they said. Despite these measures, experts stress that the episode underscores a broader threat to societal understanding, as unreliable data can distort public discourse and policy decisions.
In summary, the withdrawn church attendance survey highlights how AI and fraudulent practices are threatening the reliability of polling. As technology evolves, researchers and organizations must adapt to safeguard against misinformation, ensuring that survey data accurately reflects social trends rather than artificial constructs.
