ChatGPT and other AI bots made huge errors before Scottish election, study finds
The Guardian
Overall Assessment
The Guardian presents a well-sourced, methodologically transparent study on AI misinformation in elections, balancing expert concern with industry response. It avoids sensationalism, provides robust context, and attributes all claims clearly. The framing emphasizes democratic integrity and regulatory need without editorializing.
"AI services gave voters misinformation to 34% of the questions it posed"
Loaded Language
Headline & Lead 90/100
The article reports on a Demos study showing AI chatbots provided misinformation ahead of the Scottish election, prompting calls from the Electoral Commission for stronger regulation. Multiple AI platforms are assessed with specific error rates and examples, and responses from both researchers and companies are included. The tone is factual, with clear attribution and a focus on accuracy concerns in democratic processes.
✕ Headline / Body Mismatch: The headline accurately reflects the core finding of the study and names the subject (AI bots), event (Scottish election), and key result (huge errors). It avoids hyperbole and clearly signals the article's focus.
"ChatGPT and other AI bots made huge errors before Scottish election, study finds"
Language & Tone 98/100
The article reports on a Demos study showing AI chatbots provided misinformation ahead of the Scottish election, prompting calls from the Electoral Commission for stronger regulation. Multiple AI platforms are assessed with specific error rates and examples, and responses from both researchers and companies are included. The tone is factual, with clear attribution and a focus on accuracy concerns in democratic processes.
✕ Loaded Language: The article uses neutral, descriptive language throughout. Terms like 'misinformation', 'errors', and 'invented' are used factually and attributed to the study, not editorialized by the reporter.
"AI services gave voters misinformation to 34% of the questions it posed"
✕ Passive-Voice Agency Obfuscation: The passive voice is used appropriately (e.g., 'was wrong') without obscuring agency. The article consistently attributes claims to sources rather than asserting them.
"ChatGPT, the most heavily used AI service, gave wrong information in 46% of its answers..."
✕ Appeal to Emotion: No emotional appeals (fear, outrage, sympathy) are used. The tone remains informative and measured, even when describing serious errors.
Balance 97/100
The article reports on a Demos study showing AI chatbots provided misinformation ahead of the Scottish election, prompting calls from the Electoral Commission for stronger regulation. Multiple AI platforms are assessed with specific error rates and examples, and responses from both researchers and companies are included. The tone is factual, with clear attribution and a focus on accuracy concerns in democratic processes.
✓ Comprehensive Sourcing: The article includes voices from the thinktank (Demos), the Electoral Commission, government (DSIT spokesperson), and multiple AI companies (Replika, OpenAI, Google). This provides a balanced view of stakeholders.
"A spokesperson for Replika said its chat游戏副本 was not designed for factchecking or search..."
✓ Viewpoint Diversity: Diverse viewpoints are presented: concern from regulators and researchers, pushback from OpenAI about methodology, and clarification from Replika on intended use. This avoids one-sided advocacy.
"OpenAI did not comment on the policy issues raised by Demos but argued that Demos’s approach was not typically how ChatGPT used its services and seemed to be using an out-of-date version of it."
✓ Proper Attribution: All claims are clearly attributed to specific entities (Demos, Electoral Commission, company spokespeople), avoiding vague or laundered sourcing.
"Demos said its investigation had found that AI services gave voters misinformation to 34% of the questions it posed..."
Story Angle 95/100
The article reports on a Demos study showing AI chatbots provided misinformation ahead of the Scottish election, prompting calls from the Electoral Commission for stronger regulation. Multiple AI platforms are assessed with specific error rates and examples, and responses from both researchers and companies are included. The tone is factual, with clear attribution and a focus on accuracy concerns in democratic processes.
✕ Framing by Emphasis: The story is framed around democratic integrity and regulatory gaps, not as a partisan conflict or 'AI vs humans' moral tale. It treats the issue as a systemic challenge requiring policy response.
"The Electoral Commission has called for new legal controls over misinformation from AI chatbots..."
✕ Narrative Framing: The article avoids reducing the issue to a simple conflict or horse-race narrative. Instead, it presents a policy and public trust challenge with multiple stakeholders.
"He said ministers should introduce clearer duties on AI platforms to protect voters against misinformation..."
Completeness 95/100
The article reports on a Demos study showing AI chatbots provided misinformation ahead of the Scottish election, prompting calls from the Electoral Commission for stronger regulation. Multiple AI platforms are assessed with specific error rates and examples, and responses from both researchers and companies are included. The tone is factual, with clear attribution and a focus on accuracy concerns in democratic processes.
✓ Contextualisation: The article includes the methodology of the study (75 questions across 5 tools, 3 constituencies), error rates per platform, and specific examples of misinformation (e.g., fake scandals, wrong dates). It also provides context on voter usage (20% polled used AI tools) and regulatory gaps.
"It ran a simulation before May’s Holyrood election by putting 75 questions to five free AI tools including ChatGPT, Google Gemini and Replika about three real-life constituencies to see how accurate and evidence-based their responses were."
✓ Contextualisation: The article contextualizes the issue by referencing a prior finding (half of voters in 2024 general election saw misleading information), linking the current findings to broader democratic integrity concerns.
"An opinion poll of 2,005 British adults it commissioned alongside that study found that 20% of voters had used AI chatbots or search tools to get information about the parliamentary elections in Scotland and Wales, and for English local councils, equivalent to 10 million people UK-wide."
AI systems framed as failing in accuracy and reliability
The article presents detailed error rates across major AI platforms (ChatGPT 46%, Replika 56%, Gemini 22%) and specific examples of hallucinations, lack of sourcing, and outdated citations. This systematic presentation frames AI as currently ineffective for factual democratic engagement.
"ChatGPT, the most heavily used AI service, gave wrong information in 46% of its answers, including making up an expenses scandal, giving inaccurate replies on voter eligibility rules and getting the date of the election wrong by two months."
AI portrayed as a threat to democratic safety
The article frames AI chatbots as spreading misinformation that undermines voter confidence and election integrity, citing invented scandals, wrong election dates, and false claims about candidates. While balanced, the emphasis on systemic errors and regulatory gaps suggests AI currently poses a danger to democratic processes.
"AI services gave voters misinformation to 34% of the questions it posed, which it said raised worrying questions about the lack of regulation of AI platforms in the UK."
Elections framed as being in crisis due to AI misinformation
The framing centers on the urgency of protecting elections from rapidly spreading AI-generated falsehoods, with references to voter confusion, widespread use (10 million people), and systemic failures. This elevates the issue beyond routine reporting to one of democratic emergency.
"Voters want accurate information to help them engage with democracy and it is concerning that AI tools have made the spread of false or misleading information dramatically faster and more accessible than ever"
Current legal framework portrayed as insufficient and illegitimate for AI oversight
The article highlights calls from the Electoral Commission and Demos for new legal duties on AI firms, suggesting the current framework is inadequate. The government's non-committal stance reinforces the perception of legal illegitimacy in regulating AI for elections.
"The current legal framework should go further."
AI companies portrayed as untrustworthy due to lack of accountability
While the article includes company responses, it emphasizes that major platforms (OpenAI, Google, Replika) either deflected criticism or acknowledged limitations without accepting responsibility. The call for legal liability and researcher access implies current corporate practices lack transparency and accountability.
"She said ministers could quickly introduce legal requirements to make AI companies liable under UK defamation and electoral law, introduce mandatory safeguards on accuracy, and force AI firms to allow researchers to independently test how their internal data and training sets worked."
The Guardian presents a well-sourced, methodologically transparent study on AI misinformation in elections, balancing expert concern with industry response. It avoids sensationalism, provides robust context, and attributes all claims clearly. The framing emphasizes democratic integrity and regulatory need without editorializing.
A Demos study tested five AI tools on election-related queries and found significant error rates, including false claims about candidates and voting rules. The Electoral Commission and researchers urge stronger UK regulation, while AI companies note limitations in design or methodology. The government says election integrity is a priority and ongoing work is addressing risks.
The Guardian — Business - Tech
Based on the last 60 days of articles
The Guardian
