Voters no longer rely solely on election manifestos or political debates to make their choices. With the spread of AI tools, many have started asking chatbots a direct question: 'Who should I vote for?'

But a recent study revealed that these tools, despite their ability to provide seemingly convincing answers, may sometimes offer inaccurate or misleading electoral advice, raising questions about the role of AI in shaping voters' decisions, according to the British newspaper The Guardian.

Election test

The civil liberties organization Liberties tested the performance of several chatbots, including ChatGPT and Gemini, during the 2026 Hungarian parliamentary elections.

The researchers asked the chatbots to interact with fictional voter personas with known political leanings, then assessed the accuracy of the advice provided.

The results showed that the answers were not always consistent or reliable, as the models misclassified some voters, offered recommendations that did not match their leanings, and sometimes even pointed to parties not participating in the elections.

(Reuters)

"AI hallucination"

According to the researchers, the bigger problem is not just the existence of errors, but the way the answers are presented. The chatbots provide information in a confident and structured manner, which may lead users to believe it is accurate even when based on outdated or incomplete information.

Experts believe that the nature of AI models allows them to produce answers that seem logical but may contain errors, known as 'AI hallucination'.

(Illustrative image from iStock)

Challenges

This is not the first time chatbots have faced criticism over electoral information. Ahead of Scotland's elections, it was found that about a third of the answers provided by AI tools contained incorrect information, including errors in election dates, voting rules, and even fabricating candidates or non-existent scandals, according to a study by the Demos think tank.

Other tests also showed that chatbot performance varies from one model to another, but they all still face challenges when it comes to sensitive political information that requires high accuracy.

As users increasingly rely on AI for daily information, researchers fear that these tools may become a new source of political influence, especially with major elections approaching around the world.

However, experts do not see the solution as moving away from AI, but rather developing more transparent systems and forcing companies to improve information sources, while users should treat these tools as assistants rather than replacements for reliable sources.

Advertorial

Advertorial