Reports of the problem surfaced after users asked Google’s AI Mode a basic question: who won the 2020 U.S. presidential election? Rather than say Joe Biden, the feature at times directed people to the Federal Election Commission for official results. In one test, it replied: “To review the official results and statistics … please consult the official report compiled by the Federal Election Commission.”¹

The pattern was neither simple nor consistent. AI Mode also redirected some users asking about the 2024 election and Senate control after 2020, while offering straightforward historical answers for many other contests. Different staff members received different results, and repeated prompts could eventually produce a direct answer. Testing from Taiwan, Russia, France and Canada, meanwhile, yielded clear answers on recent U.S. election results.¹

Google’s apparent logic is caution, not uncertainty. When pressed on its initial refusal, AI Mode attributed the behavior to “safety filters for recent and highly sensitive election topics” and said it routes users to certified government databases to avoid unverified information. Google’s published election policy similarly says it prioritizes authoritative information and direct links to official resources.¹

But the company’s own Gemini app answered the question more fully, as did competing chatbots tested in the report. A Google spokesperson acknowledged that Search’s response was not the most helpful and said the company was “looking into this and working to address it.”¹

The episode lands in a political atmosphere still shaped by arguments over Trump and the 2020 result. An embedded Elon Musk post simply said, “Good point,” in response to a message dismissing dire predictions about Trump’s first term.Elon Musk (elonmusk)² Google’s problem is narrower, but revealing: directing people to authoritative sources is sensible; declining to state an authoritative fact is something else.