A Cautionary Tale About "Neutrally" Informative AI Tools Ahead of the 2025 Federal Elections in Germany

AI tools claiming neutrality may harbor significant political biases: Large language models tested in this study showed strong alignment with left-wing parties (over 75%) while aligning with right-wing parties only around 30%, demonstrating that these systems are not the objective information sources they may appear to be.

Voting advice applications contain substantial factual errors: Two AI-based voting tools designed to objectively match voters with parties showed deviations from actual party positions in 25-50% of cases, meaning voters could receive misleading information about where parties actually stand on key issues.

AI systems can be manipulated to spread false political claims: The study found that simple prompt injections caused one VAA to generate severe hallucinations, including fabricating non-existent connections between political parties and right-wing extremist groups—a particularly concerning finding ahead of elections.

"Neutral" tools require independent verification before deployment: These findings suggest that AI-based political information systems need rigorous fact-checking and bias audits before being used to inform voters, rather than being assumed reliable simply because they claim objectivity.

Voters should be cautious about relying solely on AI for political guidance: Given the documented biases and errors, citizens should cross-reference AI voting tools with official party statements and multiple information sources rather than treating algorithmic recommendations as definitive.