**
A recent study has raised serious alarms about the reliability of AI chatbots in providing electoral advice, following Hungary’s parliamentary elections in April. Researchers found that popular chatbots, including ChatGPT and Gemini, failed to accurately recommend political parties, often suggesting parties that were not even running. This troubling conclusion underscores the potential risks of using AI for political guidance during elections.
Inaccurate Recommendations Unveiled
The investigation, conducted by the civil liberties organisation Liberties, highlights a concerning trend: AI chatbots frequently misclassified voter profiles, omitted relevant parties, and even recommended parties that were absent from the ballot. In a striking example, when users presented profiles aligned with the newly dominant Tisza party, the chatbots failed to recommend it 90% of the time. Instead, they often suggested smaller parties that were unlikely to meet the 5% threshold necessary to secure parliamentary representation.
The study’s findings are particularly significant given the political landscape shift in Hungary, where Péter Magyar’s Tisza party recently unseated Viktor Orbán and his long-standing Fidesz party. While Tisza emerged victorious in a landslide, the researchers caution that in a more competitive electoral setting, the inaccuracies of AI guidance could have severe implications for voter decision-making.
Consistency Issues and Opacity
One of the most alarming aspects of the research was the inconsistency of the AI systems. Identical voter profiles yielded radically different recommendations upon successive queries. In fact, both ChatGPT and Gemini included parties not participating in the 2026 elections in 96% of their responses. This raises questions about the reliability of AI-generated electoral advice, which often appears authoritative and persuasive, despite its lack of transparency.
The study noted that while these AI models began with disclaimers about their inability to provide political advice, they then proceeded to deliver detailed recommendations, further complicating users’ ability to discern the accuracy of the information. The researchers expressed concern that users might mistakenly trust these outputs, as they appeared well-argued and precise on the surface.
Gaps in Regulation Highlighted
This report shines a spotlight on a critical regulatory gap concerning the use of AI in electoral contexts. Although the EU’s AI Act requires general-purpose AI systems to assess systemic risks, and the Digital Services Act addresses electoral integrity, chatbots currently slip through the cracks. Liberties advocates for stricter safeguards to ensure that AI systems offering political advice are transparent, accurate, and accountable.
Eva Simon, head of Liberties’s tech and rights programme, emphasised the need for reform: “General-purpose AIs should not present opaque and unstable political matching as if it were reliable electoral guidance for voters. Democracy cannot rely on opaque systems that claim neutrality while delivering advice they cannot explain, reproduce or guarantee to be accurate.”
Why it Matters
The implications of this study extend beyond the borders of Hungary. As AI technology becomes increasingly integrated into our daily lives, especially in critical sectors like politics, the reliability of these systems comes into question. If voters cannot trust AI-generated electoral advice, it undermines the very foundations of democratic participation. Ensuring that AI systems are rigorously regulated and capable of providing accurate, consistent information is essential for maintaining the integrity of electoral processes worldwide. The future of democracy may well hinge on how we tackle these challenges today.
