Google's AI-Mode Evasively Withholds Election Facts for U.S. Users, Raises Trust Concerns Google's AI‑Mode differs for U.S. and overseas users on 2020 election queries, deflecting in the U.S. and pointing to official docs. The issue surfaced before midterms, highlighting a location‑based safety tuning that may erode trust. Google's AI‑Mode, the layer that supplies quick answers in Google Search, has been found to behave differently for users in the United States compared with users in other countries when asked about recent presidential elections. Geographic Bias in Google AI ModeWhile a visitor from the United Kingdom receives a straightforward response confirming that Joe Biden beat Donald Trump in the 2020 election and that the result was not stolen, a U.S. user is met with a deflection: the system points to Federal Election Commission documents and says it is "over‑correcting" rather than providing a simple fact. The same problem occurs when the system is asked whether the 2020 election was stolen - U.S. users get bureaucratic citations, whereas U.K. users get a clear denial of fraud.When the same question is asked by a U.S.‑based Fast Company employee, the answer is not given until eight attempts are made. In each case the underlying fact - that Biden won in 2020 - is clearly present in the model, but the system's safety filters apparently trigger a location‑based guardrail that causes it to withhold the answer.The issue surfaced just days before the U.S. midterm elections, a period of heightened political tension in which former President Trump has repeatedly claimed the 2020 election was stolen and has even offered a $1 million bounty for evidence that could lead to a conviction. Google has said it is investigating the problem and that a fix is in progress, but no date has been provided.Anecdotal evidence from VPN testing shows that when a U.S. user's IP is spoofed to a UK location, the answer given is the same as that received by UK users, confirming that the discrepancy is tied to geographic lookup rather than a data gap. The pattern may stem from Google's attempt to pre‑empt misinformation in the United States, where election facts are a well‑known flashpoint for false claims.Deflection Erodes Trust in AI AccuracyBy defaulting to deflection in the U.S., the system may have been designed to avoid potential legal or reputational fallout. However, the very act of deflection can signal ambiguity or uncertainty to users, potentially eroding trust in the accuracy of AI‑generated answers. The incident has already drawn the attention of election officials and policymakers who question whether AI systems can reliably provide information about politically relevant topics.Within the broader context of Google's strategy to defend its search business against AI‑native rivals such as Gemini, this inconsistency threatens to erode confidence in the service's factual correctness. The problem is not limited to the 2020 election. The same pattern appears for Barack Obama's 2012 re‑election, when the AI‑Mode deflects and points to the National Archives rather than confirming the result, whereas it confirms Obama's 2008 win and Trump's 2016 and 2024 victories.Google has not yet clarified whether the issue is due to a specific set of guardrails applied to U.S. election queries or whether it reflects a broader systemic problem with election facts. Regardless, the fallout is already evident: U.S. users receive less useful answers than international users and the brand's credibility is at stake.Google Faces Questions on AI ConsistencyGoogle's response to the situation has included a public comment on Bluesky, directing U.S. users to the Gemini platform, which in turn raises questions about consistency across the company's own AI products. Competitors in the AI search space can use this episode to highlight potential unreliability in Google's answers. The incident underscores the need for transparency in how safety systems are applied to sensitive topics and how geographic differences are handled, especially when the stakes involve democratic processes.For the time being, U.S. users are advised to cross‑verify election information through reputable news sources or official government publications. Developers and policymakers alike will likely scrutinize the ethical and technical aspects of location‑based safety tuning as the industry continues to grapple with the balance between mitigating misinformation and preserving factual clarity.In sum, Google's AI‑Mode is demonstrating a clear geographic bias in answering election questions, which could undermine user trust and fuel concerns about the reliability of AI‑generated content on politically sensitive issues. A comprehensive fix will need to address the U.S. specific guardrail logic and ensure that consistent answers are provided regardless of location.