‘Our assumptions are broken’: how fraudulent church data revealed AI’s threat to polling
TL;DR
A 2024 Bible Society report claimed church attendance in Britain was surging – but the underlying data was fraudulent. Paid survey participants use AI tools to generate fake responses at scale, collecting payments without genuine input. Experts warn that automated bots and LLMs are systematically corrupting online surveys across industries. The problem extends far beyond religious studies – political polling and market research face the same threat.
Nauti's Take
This is not a fringe problem caused by a handful of dishonest clickworkers – it is a structural attack on the data foundations of entire industries. When LLMs produce survey responses convincing enough to pass quality checks, trust in quantitative social research breaks down at a fundamental level.
The irony is almost too rich: the same AI technology marketed as a tool for uncovering truth is here industrially manufacturing falsehoods. The research industry urgently needs new verification standards – before the next polling disaster makes headlines.
Briefingshow
Opinion polls are a cornerstone of democratic decision-making and market research – when their data is corrupted by AI-generated fake responses, politicians, media, and businesses lose a critical source of truth. The Bible Society case demonstrates this is already a real, publicly visible problem, not a hypothetical risk. Most alarming: the fabricated data was initially celebrated and widely circulated as evidence of genuine social trends.
Sources