Anthropic Opens AI-Experience Study With an Option to Publish Full Interviews

Claude users can now tell Anthropic what they want from AI—and choose to let everyone else read the conversation. Anthropic opened its new study on September 29, running through October 6, 2026. The optional release of complete interviews gives outsiders access to participants’ own words, but the company warns that even without names, those words may reveal who spoke.

The project follows a December 2025 study that Anthropic says drew responses from 81,000 people. That effort made high-level findings and a small selection of quotations public. According to the company, it helped shape the Anthropic Institute’s agenda and continues to guide its societal-impact and economics research.

The interview method is not new. Anthropic introduced Anthropic Interviewer in December 2025 as a Claude-powered tool that conducts interviews automatically and feeds results to human researchers. This follow-up updates the questions, opens participation across more Claude products and adds the public-sharing choice.

A short interview, three broad questions

The study is open to Free, Pro and Max users on Claude and Claude Code whose accounts are at least two weeks old. Each interview takes roughly 15 minutes, with the AI interviewer asking the questions. Anthropic is seeking experiences and expectations rather than answers to a single policy proposal.

• Which experiences with AI have mattered most, whether positive or negative?
• What would participants like AI to change about areas such as work, school, healthcare or government?
• What do participants want from the companies developing AI, including Anthropic?

Anthropic says public transcripts will let researchers and policymakers examine responses across regions and languages, pursue questions its own staff did not ask, and check whether its published findings reflect what participants actually said. Publication would also preserve criticism of AI companies in a record others can reference.

We can remove our copy of your interview on request, but we can’t delete what people have already saved.

Anthropic, study FAQ

The public choice comes after review

Participants can reread their full interview before deciding whether to publish it. Anthropic says it explains the risks before the interview, shows examples of sensitive information to avoid during it, and confirms the sharing decision afterward. A public transcript will include the country associated with the conversation, but not account details such as a name or email address.

That is not a guarantee of anonymity. Anthropic warns that details such as a job, city, age or past project can be combined to identify someone, including with AI. Public interviews could be read by employers, relatives, governments or scammers. The company advises participants not to share information that could identify themselves or others.

Keeping an interview out of the public release does not keep it out of the research. Anthropic will still analyze it and may share aggregated or de-identified findings. It says results will inform its models and products. Participants who stop early using the “End interview” button will instead have their partial interview permanently deleted, without analysis or publication.

An open record, not a public-opinion sample

The interviews are due online after the study and initial analysis are complete. Anthropic does not plan to edit them, though a final review may exclude transcripts containing confidential information or violating its policies. The public record will therefore contain consenting participants’ words, subject to that publication review.

Its reach has another limit: all participants use Claude. Anthropic acknowledges that they do not represent the public, and those opting into publication may differ from those who decline. The interview questions also shape what people share. The next release will offer accounts of AI’s impact—not a representative measure of what everyone wants from it.