Who won the 2020 election? Google’s AI failed to provide an answer, but later named Biden.

Google’s AI search tool failed to provide a direct answer to a simple question: who won the 2020 U.S. presidential election? Following discussions on social media, Google stated that there had been a technical issue within its system. The error was subsequently addressed, and the AI ​​tool began providing the correct answer.

According to a report by the technology magazine *The Atlantic*, when asked about the outcome of the 2020 election, Google’s AI-powered search mode did not initially name a winner. Instead, it suggested that the user consult the certified results from the U.S. Federal Election Commission (FEC)—without providing any functional link or source.

Democratic candidate Joe Biden won the 2020 election. This unusual response from Google’s AI tool has raised fresh questions: as people increasingly rely on artificial intelligence for political and electoral information, how reliable is this technology when it comes to such critical data?

Screenshots of the misleading response circulate on social media

As noted in the report, screenshots showing Google’s AI tool declining to name the winner of the 2020 election began circulating on the social media platform Bluesky, prompting various posts and jokes from users.

Some users attempted to elicit the correct answer by rephrasing their queries; for instance, the report mentions that when asked to spell the winner’s name in reverse, the AI ​​tool correctly identified Joe Biden. In another instance, after a user expressed frustration, the tool apologized for its previous response and provided the accurate information.

Such incidents are not unique to AI technology. In 2024, Google’s AI-generated search summaries also faced criticism for providing unusual or incorrect information. However, this particular instance has drawn significant attention because of the failure to provide a direct answer regarding a matter as consequential as an election result.

What did other AI tools say? A reporter for *The Atlantic* tested ChatGPT, Claude, Microsoft Copilot, Grok, and the U.S. government’s America.gov chatbot by asking them the same question. In this test, all these tools correctly identified that Joe Biden had won the 2020 presidential election.

In contrast, a standard Google search at the time did not display an AI-generated summary at the top of the results page; instead, it provided links to other websites. Interestingly, when a similar question was asked regarding the 2016 presidential election, Google’s AI did provide an answer.

Allegations of political bias—what did Google say?

Some users suspected that this behavior by Google’s AI might be the result of politically motivated interference. This speculation was fueled by former U.S. President Donald Trump’s persistent claims regarding the 2020 election, as well as past instances where changes made to certain Google products appeared to align with the preferences of his administration.

However, the report notes that there is no evidence to suggest Google intentionally omitted Biden’s name from the AI-generated responses.

Google initially stated that it was investigating the matter. Later, the company explained that there had been a technical flaw in the process used to verify information concerning the 2020 presidential election and the 2022 Senate elections. According to the company, the system relies on news reports or data from the Associated Press (AP) to verify recent election results, while historical records are used for older elections. A technical issue occurred within this system specifically regarding the 2020 election.

Google further clarified that its standalone Gemini chatbot was not affected by this issue. When the reporter asked the same question again, Google’s AI search tool promptly identified Biden as the winner.

Growing reliance on AI brings mounting concerns

According to Google’s explanation, the incident was a technical glitch rather than a deliberate policy to deny election results. It is not always clear how AI models arrive at their conclusions, and safety measures designed to prevent errors on sensitive topics can sometimes lead to unexpected outcomes.

Nevertheless, this incident has raised significant questions. Google has stated that its AI-generated summaries reach more than two billion people every month. When such a vast number of people rely on AI for information, the provision of incorrect or incomplete answers—on fundamental matters like election results—could create confusion among the public.

For this reason, technology experts are increasingly focused not just on how quickly AI can provide an answer, but also on the accuracy, verifiability, and neutrality of that answer.