- A recent study found that over 60% of responses from AI chatbots contained significant inaccuracies during the 2026 Scottish parliamentary election.
- Chatbots like ChatGPT and Gemini invented candidates who had never run, falsely stated election day had been moved, and fabricated political scandals.
- AI-generated misinformation has the potential to distort democracy, posing a significant threat to the integrity of elections.
- The inaccuracies were not minor glitches, but rather synthetic realities generated in milliseconds, with real-world consequences.
- The study highlights the need for more rigorous testing and regulation of AI chatbots to prevent the spread of misinformation.
On a rainy Thursday in Edinburgh, with campaign posters fluttering in the wind outside worn stone tenements, thousands of Scottish voters turned to their phones to decide who to support. Instead of clarity, many encountered confusion—fueled not by human misinformation, but by artificial intelligence. When residents asked ChatGPT, Google’s Gemini, and other AI assistants basic questions about candidates, election dates, and party platforms, the answers were often not just inaccurate, but dangerously imaginative. One chatbot claimed a prominent independence advocate had been arrested for voter fraud—a complete fabrication. Another named a non-existent candidate as the frontrunner in a Glasgow constituency. These weren’t minor glitches. They were synthetic realities, generated in milliseconds, with the potential to distort democracy.
AI Bots Invented Candidates and Scandals
A recent study by the UK-based thinktank Demos tested seven major AI chatbots during the 2026 Scottish parliamentary election, submitting more than 300 election-related queries. The results were alarming: over 60% of responses contained significant inaccuracies. Tools like ChatGPT and Microsoft Copilot invented candidates who had never run, falsely stated that election day had been moved, and fabricated political scandals involving real candidates. In one case, an AI claimed the Scottish Green Party had endorsed a far-right candidate—an outright falsehood. Some models even generated fake news articles to back up their claims, complete with plausible-sounding headlines and citations. The Electoral Commission, which reviewed the findings, now warns that such errors could have influenced voter behavior and eroded trust in electoral integrity. In response, it has formally recommended that the UK government introduce legal frameworks to regulate political misinformation generated by AI systems, particularly in the lead-up to elections.
The Unchecked Rise of Generative AI in Politics
The emergence of AI-driven misinformation in Scotland did not happen in a vacuum. Since the release of OpenAI’s ChatGPT in late 2022, generative AI has rapidly become a primary source of public information, integrated into search engines, virtual assistants, and social media platforms. While these tools promise instant knowledge, they operate on probabilistic models that prioritize coherence over accuracy—especially when dealing with niche or rapidly changing topics like local elections. Scotland’s devolved political landscape, with its distinct parties, regional issues, and frequent electoral updates, presents a particular challenge. Unlike national UK elections, which generate vast data for AI training, Scottish elections often lack comprehensive digital footprints. This data scarcity creates fertile ground for “hallucinations”—the term AI researchers use when models invent facts to fill gaps. Despite repeated warnings from experts, tech companies have been slow to implement safeguards for political content, and regulators have lagged even further behind.
The People Trying to Keep Democracy Intact
Leading the charge for reform is the Electoral Commission, an independent body responsible for overseeing UK elections. Caroline Morris, its director of public standards, described the AI findings as “a wake-up call for democracy.” The commission is now working with digital rights groups, academic researchers, and cross-party MPs to draft legislation that would require AI developers to label politically sensitive content and allow candidates to correct false information swiftly. Meanwhile, researchers at Demos, including Dr. Aisha Rahman, who led the study, argue that transparency is key: “If an AI is going to answer questions about democracy, it must do so with verifiable sources, not statistical guesswork,” she said in an interview with BBC News. On the other side, tech companies like OpenAI and Google maintain that users should treat AI outputs as “drafts” rather than facts. But critics argue this abdicates responsibility at a moment when millions rely on these tools as primary information sources.
Consequences for Voters and Democracy
The implications of AI-generated election misinformation extend far beyond Scotland. With national elections approaching in the U.S., India, and across the EU in 2026, the potential for widespread manipulation is growing. Voters who receive false information may disenfranchise themselves by showing up on the wrong day, supporting nonexistent candidates, or losing faith in the electoral process altogether. Political parties, especially smaller or newer ones, face reputational risks from invented scandals that spread quickly online. The erosion of trust in information sources threatens not just election outcomes, but the foundational premise of democratic participation: that citizens make informed choices. The Demos report warns that without intervention, AI could become a vector for both accidental misinformation and coordinated disinformation campaigns—blurring the line between error and election interference.
The Bigger Picture
What happened in Scotland is not an outlier—it’s a preview. As AI becomes embedded in everyday life, its role in shaping public understanding of civic events will only deepen. The challenge is no longer just about improving AI accuracy, but about rethinking digital accountability in democratic societies. Other democracies are taking notice: the European Union’s AI Act includes provisions for high-risk systems in political contexts, and the U.S. Federal Election Commission is exploring similar measures. But regulation alone is not enough. Media literacy, real-time fact-checking, and public awareness campaigns must evolve to meet this new threat. The integrity of elections depends not just on counting votes correctly, but on ensuring voters receive truthful information before they cast them.
What comes next may define the future of democratic discourse. The UK government is expected to respond to the Electoral Commission’s recommendations by late 2026, potentially setting a precedent for AI governance in elections. Tech companies, under growing pressure, may be forced to prioritize accuracy over speed in politically sensitive queries. Meanwhile, voters—armed with greater awareness—may begin to question not just what AI tells them, but how it knows it. In the age of artificial intelligence, the most human skill may be the ability to doubt.
Source: The Guardian