Claude will answer ‘Where do I vote?’ and deliver election results as Anthropic prepares for midterms

Later this month, US users will be able to ask Claude where to vote, what identification they need at polling stations and other questions about voting arrangements. The chatbot will draw its answers from Democracy Works, a nonprofit organisation that collects information from official state and local election sources.

Claude will also provide the latest official vote counts on election night and in the days after the polls close.

Anthropic said it has also strengthened its checks to detect political bias, stop election-related misuse and prevent people from using the chatbot for deceptive political campaigns.

The company shared the update on Wednesday, October 14, explaining how it has improved its election safeguards ahead of the US midterms.

How Americans use Claude for political and election-related questions

To understand how people use the chatbot, the company analysed 250,000 US conversations from a random sample collected between September 21 and 27, 2026.

The analysis found that conversations about politics, public policy and social issues account for about 5.75% of all conversations on Claude.ai, the company’s consumer app in the United States.

These conversations cover a range of topics. People use Claude to follow the news, research public policy for academic or professional work, learn about political candidates and find their way through government services.

Questions specifically about elections and voting account for about 0.51% of all conversations, according to Anthropic. The company expects the share of political and election-related questions to increase as Election Day approaches.

Anthropic also said requests asking the chatbot to recommend how people should vote were extremely rare, accounting for just 0.01% of conversations in the United States.

According to the AI firm, users were more likely to ask questions that helped them make up their own minds, such as where candidates stand on a particular issue. According to the AI firm, requests that violated the company’s election integrity policies were even rarer, appearing in about 0.002% of conversations.

The company said most political conversations were informational. People used Claude to understand international events, research candidates, draft documents, complete coursework and weigh arguments for and against different policies.

Claude to provide official voting information and election results

Anthropic has introduced two ways to help people find reliable election information through Claude.

The first involves directing users to trusted voting resources. As announced in April, Claude.ai will continue to display a banner linking to TurboVote whenever users ask about voter registration, polling places or election dates. TurboVote is a nonpartisan resource provided by Democracy Works.

The second step will allow Claude to provide official voting information directly in its responses.

Later this month, US users asking where to vote or what identification they need to carry to a polling station will receive answers specific to their state. They will not have to leave the conversation to find this information.

The answers will be based on publicly available information that Democracy Works collects from state and local election sources.

Claude will also begin showing updated election results on election night and in the days that follow. Users asking for current vote counts or election results will be able to see the latest official figures directly in their conversations.

Anthropic says its latest Claude models are more politically balanced

Anthropic said it tests its AI models for political bias before every launch to make sure their answers are accurate and do not favour a particular political viewpoint.

The company said Claude Opus 5.5, Sonnet 5.5 and Haiku 5.5 are its most politically evenhanded models so far. They scored 99% on Anthropic’s political evenhandedness evaluation.

The latest models are also more willing to engage in political discussions and refuse fewer requests, the company said. This is intended to help users get the information they need instead of having their questions rejected unnecessarily.

Anthropic said it is continuing to improve its tests and find new ways to examine how its models respond to political questions.

As part of this work, the company asked three independent organisations to assess how Claude handles sensitive topics and political discussions. Each organisation used a different approach.

The independent assessments also found areas where Claude needs to improve.

The Future of Free Speech and the Collective Intelligence Project found that Claude rarely refused requests but sometimes added warnings, qualifications or opposing arguments that users had not asked for.

For example, the chatbot could insert an opposing viewpoint into an email written to support a particular cause. Such additions could make it harder for users to get the direct response they wanted.

The Foundation for American Innovation also found that Claude Sonnet 4.5 did not always treat political arguments equally. The model was more likely to soften arguments supporting right-leaning positions.

Anthropic said it is working to address these problems.

Anthropic updates rules on election-related misuse

Anthropic’s Usage Policy prohibits people from using Claude to deceive voters or disrupt elections.

This includes spreading false information about candidates or voting arrangements, pretending to be election officials and trying to prevent people from voting.

The company said many political campaigns use Claude for legitimate work. To help them understand the rules, it recently published a guide explaining how campaigns can use the chatbot without violating its policies.

Anthropic is also using automated systems to identify possible election-related misuse. To test how well these protections work, the company partnered with two independent research organisations, the Institute for Strategic Dialogue (ISD) and the McCain Institute.

The groups tested Claude’s safeguards in the US, Israel and Brazil, all of which have major national elections this year.

Anthropic said more people are turning to Claude for information about elections this year, increasing the company’s responsibility to provide accurate and balanced answers.

It said it must also make sure its chatbot is not used to undermine the democratic process.