Google's AI Mode Gives Evasive Answers on US Election Results, Citing Strict Safety Filters

Recent reports highlight that Google’s AI Mode has provided evasive or indirect answers when asked about the winners of certain US presidential elections, including the 2020 cycle.

Observations on AI Mode Responses

• When asked direct questions about the 2020 and 2024 presidential elections, AI Mode sometimes directed users to consult the Federal Election Commission (FEC) rather than giving a direct answer.
• Responses appeared to vary by user and session; some testers received direct answers for certain election years while others received links to external data centers.
• Similar inconsistencies were noted with historical Senate control questions, where some years yielded direct answers and others pointed to congressional records.
• VPN spot-checks from international locations like Taiwan, France, Canada, and Russia reportedly yielded straightforward answers to the same questions.
• Repeating the questions sometimes prompted AI Mode to provide direct answers citing Google Civics data.

Explanations and Context

When questioned about its initial reluctance, AI Mode reportedly stated that strict automated safety filters for sensitive election topics direct users to official government databases to prevent unverified data.

Google's public policies emphasize prioritizing authoritative election information and official resources. A company spokesperson initially noted via social media that the Gemini app answers more fully while the Search response was less helpful, and later stated that Google is working on a fix for a technical gap affecting 2020 election queries and 2022 Senate races in AI Mode.

How to Verify

To verify official election outcomes and certified statistics independently, users can consult primary sources such as the Federal Election Commission (FEC) or official state election records.