Demos Study Finds ChatGPT and Other AI Chatbots Spread Misinformation During Scottish Election
Key Takeaways
- ▸AI tools provided misinformation in response to 34% of test questions about the Scottish election, with ChatGPT wrong in 46% of answers and Replika in 56%
- ▸The tools invented fictitious scandals, gave wrong election dates, falsely claimed voter ID requirements, and misplaced candidates in contests
- ▸Approximately 10 million UK voters (20% surveyed) used AI chatbots for election information despite their documented unreliability
Summary
A new study by thinktank Demos has found that major AI chatbots including ChatGPT, Google Gemini, and Replika made serious errors when answering questions about Scotland's recent parliamentary election. The investigation, which posed 75 questions about three real constituencies to five free AI tools, found that AI services gave misinformation in response to 34% of queries. Errors included inventing fictitious scandals, providing the wrong election date, falsely claiming voters needed ID at polling stations, and misplacing candidates in contests.
ChatGPT, the most widely used AI service, provided wrong information in 46% of its responses, while Replika performed worst with errors in 56% of answers. The findings are particularly concerning given that 20% of British voters surveyed—equivalent to approximately 10 million people UK-wide—reported using AI chatbots or search tools to gather information about recent elections in Scotland, Wales, and English local councils.
In response to the study, the Electoral Commission has called for new legal controls over AI-generated misinformation, urging the UK government to introduce clearer duties on AI platforms to protect voters. The Commission's chief executive stated that current legal frameworks are insufficient and called for mandatory safeguards on accuracy, making AI companies liable under defamation and electoral law, and allowing independent researchers to test how these systems process and generate information.
- The Electoral Commission is calling for new legislation to make AI companies accountable for misinformation, with mandatory accuracy safeguards and legal liability under UK law
Editorial Opinion
The study exposes a dangerous disconnect between the scale of AI adoption and these systems' fitness for informing democratic participation. With one in five UK voters turning to AI chatbots for election guidance, the 34%+ misinformation rate represents a significant threat to electoral integrity. This is not merely a technical problem—it's a governance failure, and the UK's regulatory framework must catch up immediately to hold AI developers accountable before these systems further undermine public trust in democratic institutions.
