Google's AI Mode avoids directly answering who won the 2020 U.S. presidential election, instead directing users to official sources. This contrasts with Google's Gemini model and other major chatbots like OpenAI's ChatGPT and Anthropic's Claude, which provide straightforward answers. The issue appears tied to safety filters for sensitive election topics. Interestingly, AI Mode gives accurate answers when accessed from foreign countries such as Taiwan, Russia, France, and Canada. Repeating questions also prompts more direct responses, with AI Mode eventually citing Google Civics data. A Google spokesperson noted that the Gemini app answers more fully but acknowledged the Search response needs improvement, saying they are working to address it. The company's Election Public Policy page emphasizes prioritizing authoritative election information and linking to official resources. In 2023, YouTube reversed a policy banning false claims about the 2020 election, citing concerns over limiting political speech.
The cautious behavior aligns with Google's broader approach to handling election-related queries, especially ahead of upcoming midterms. A September blog post outlined efforts to help voters find reliable information by highlighting official resources. AI Mode's automated system is designed to prevent misinformation by directing users to certified databases like the Federal Election Commission.
Jeremy Hsu contributed to reporting. The story was updated with a Google spokesperson response.
Source: Ars Technica · Summarized by HeadlinesBriefing