‘Our assumptions are broken’: how fraudulent church data revealed AI’s threat to polling

If you had been keeping tabs on the news about church attendance in Britain lately, you would be forgiven for thinking the country was in the midst of a Christian revival.Stories of swelling congregations, filled with young people returning to the flock, spurred on by everything from social media to a rise in bible sales appeared to be confirmed by a 2024 report from the Bible Society.Based on data collected by a YouGov survey, it claimed church attendance was increasing in England and Wales. The findings drove headlines, and the narrative was established.There was just one problem – the survey turned out to be based on “fraudulent” data and has been withdrawn. And academics and experts are warning that this episode should serve as a parable, not about a renaissance in religion, but of the false prophets of artificial intelligence.Researchers have said online opt-in surveys are becoming increasingly infested with bogus data as respondents who are often paid for their participation use AI to fill in questionnaires at speed.These particular surveys are self-selecting research forms that can shape national discourse. But a major fault, experts say, is that they are susceptible to “survey farmers”, and this vulnerability means the results should be treated with caution when trying to understand social trends.David Voas, a quantitative social scientist and emeritus professor at University College London said: “This sort of information, misinformation, is just very difficult to correct once it starts spreading. 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.“We are finding our confidence in these sorts of polls undermined, and then it’s very difficult to move back,” he added.Voas said the problem was a general one, not just isolated to YouGov.“It’s a growing problem because if you [as a participant] can work at scale, you can actually generate a reasonable amount of revenue, even in western terms, never mind global south terms,” he said.According to a 2026 report, YouGov uses a random sample method which means participants cannot choose which survey they want to complete. A YouGov spokesperson told the Guardian: “We select which surveys to send them, so there is no way of joining to influence the result of a piece of research.”A broken survey modelThe growth of AI has exacerbated the problem. Sean Westwood, an associate professor in the department of government at Dartmouth College, in New Hampshire, said: “The assumption with survey research – that someone gives coherent, logical answers, they’re a real person – that assumption is now broken.”There is no evidence to suggestAI use was the source of the fraud in the church attendance numbers recorded by YouGov in 2024. However, Westwood said AI had the potential to influence online survey research. “The tools to do all of this are cheap, accessible and available right now.”He said AI models posed an existential threat to our understanding of society. “The agent can be weaponised. A single sentence of instruction is enough to systematically bias its answers on political polls or geopolitical questions, while keeping its demographic profile intact so the manipulation is invisible to standard screening,” he said.“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 rate at which AI is being used to complete these surveys is difficult to detect. “We don’t know the precise scope, and that’s part of the problem,” said Westwood.And, he added, the rapid transformation in AI technology makes it difficult to combat its potential influence. “A researcher might design a clever new trap that catches today’s models, but model development moves so fast that the fix is likely obsolete within months.”The youth influenceThe report by the Bible Society claimed that church attendance was increasing at the highest rate among young people in England and Wales. Courtney Kennedy, the vice-president of methods and innovation at Pew Research Center, said opt-in estimates for people under the age of 30 tend to contain high levels of error and are more likely to stem from “click farms”.“In general, people who are highly skilled using the internet and concealing their identity skew younger,” said Kennedy. “Bogus cases want to qualify for as many surveys as possible. It is well known in the industry that young adults are hard to reach for surveys. So from this standpoint, it is advantageous to self-present as young because surveys tend to need such respondents.”Kennedy added: “Bogus respondents tend to respond in the affirmative, no matter what is asked. This is called positivity bias, which inflates the estimate.”Voas said the problem with the Bible Society report was not only the fraudulent respondents but the failure to critically compare the findings of the YouGov survey with other available research by the churches themselves.“If you were doing serious scholarly research, you would need to review the literature and see what other evidence was out there,” said Voas.A YouGov spokesperson said: “The rise of organised survey farms, bots, and now AI-assisted responses makes detection a vital, continuous and constantly evolving discipline. YouGov uses identity checks, device fingerprinting, multi source geolocation, real-time threat scoring, and payout oversight so bad actors do not slip through the net.“When someone joins the YouGov panel we link the information they supply with every data point we can observe about their device, location and behaviour. From this we decide who to invite, who to verify and, if necessary, who to remove.”