Anthropic has opened a study using Anthropic Interviewer to gather Claude users’ experiences and expectations of AI, with an optional route to publish their complete interview and associated country online. For eligible participants, the roughly 15-minute interview offers a way to describe AI’s benefits, harms and desired changes; opting into public release, however, can expose personal details in a record that cannot be fully withdrawn.
The company says the study asks about participants’ most meaningful positive and negative encounters with AI, changes they would like AI to support in areas including work, school, healthcare and government, and what they want from AI developers, including Anthropic. Anthropic will analyze the interviews through its Societal Impacts research, publish findings and use results to inform model and product development.
The study runs from September 29 through October 6, 2026. It is available to Free, Pro and Max users of Claude and Claude Code whose accounts are at least two weeks old. Anthropic says participants can end an interview at any time; if they do, the partial interview is permanently deleted and will not be analyzed or published.
Public release is not required. A participant who chooses it will have their full interview and the country associated with the conversation published, while Anthropic says it will exclude Claude account information such as name and email address. The company says it does not plan to edit interviews before release, although it will conduct a final review after the study and may withhold interviews it considers irresponsible to publish, including those containing confidential information or material that violates its policies.
Anthropic presents the public option as a way to make the underlying testimony available beyond the company. It says researchers, policymakers and other readers could examine interviews across regions and languages, pursue questions outside Anthropic’s analysis, and assess whether the company’s published findings reflect what participants actually said. The interviews are due to be published online after the study and Anthropic’s initial analysis are complete.
The design changes the level of source material available compared with Anthropic’s December 2025 study, in which 81,000 people shared hopes and worries about AI. Anthropic says that earlier work informed the Anthropic Institute’s agenda and its ongoing societal-impacts and economics research, but only high-level findings and a small selection of quotations were released publicly. The new study uses updated questions, covers more Claude products and, for the first time, lets participants elect to make an interview public.
Public record, limited anonymity
The publication choice comes with a substantial qualification: removing an account name and email address does not guarantee anonymity. Anthropic warns that details such as age, job, city, education, employer or past projects may enable someone to connect a response to its author. It also says researchers have shown that re-identification can work with Anthropic Interviewer interviews when small details are combined, potentially with AI.
Participants are shown examples of information Anthropic advises them not to disclose, can review their whole interview before choosing whether to publish it, and receive a final confirmation if they opt in. Even so, the company stresses that public interviews can be read, saved and used by anyone, including employers, friends, relatives, governments, companies, scammers and others who may misuse personal information. Anthropic can remove its own copy at a participant’s request, but cannot erase copies that others have saved; it advises treating public release as permanent.
Participants who do not choose public release are still part of the study unless they end the interview early. Their complete interviews will not be published, but Anthropic says they may contribute to aggregated or de-identified findings. That distinction means the public archive will capture only the subset of respondents willing to expose their accounts publicly, rather than all study participants.
What the results can—and cannot—represent
Anthropic explicitly cautions against treating the study as a representative measure of public opinion. Every participant is already a Claude user, and the people who authorize publication may differ from those who do not. The company also notes that interview questions shape what respondents choose to discuss. Those limits apply both to the published interviews and to any conclusions drawn from the dataset.
The study is therefore a structured collection of views from a defined product-user population, not a population survey or an assessment of AI performance. Its prospective value is in preserving participants’ accounts and making an opt-in subset available for outside examination, while its privacy protections require participants to avoid identifiers themselves. Source: Anthropic.
Definition. The study is an opt-in collection of Claude users’ AI experiences and expectations, with an optional public release of complete interviews and associated country.
| December 2025 study | New study |
|---|---|
| 81,000 people shared hopes and worries about AI | Eligible Claude users share experiences and expectations through Anthropic Interviewer |
| Only high-level findings and a small selection of quotations were released publicly | Participants may elect to make a complete interview public |
| Earlier work informed the Anthropic Institute’s research agenda | Anthropic says results will inform societal-impacts research and model and product development |
Key takeaways
- The study runs from September 29 through October 6, 2026, for eligible Free, Pro and Max Claude and Claude Code users with accounts at least two weeks old.
- Participants can stop an interview at any time; Anthropic says partial interviews are then permanently deleted and not analyzed or published.
- Public release is optional; published interviews include the full conversation and associated country, but exclude Claude account name and email address.
- Removing account details does not ensure anonymity because personal details in an interview may enable re-identification.
- People who decline public release may still contribute to aggregated or de-identified findings.
- Anthropic cautions that the study is not representative of public opinion because participants are Claude users and public-release volunteers may differ from other respondents.
FAQ
Who can participate in Anthropic’s study?
It is available to Free, Pro and Max users of Claude and Claude Code whose accounts are at least two weeks old.
Is publishing an interview required?
No. Participants can take part without authorizing public release; their complete interviews will not be published.
What is published if a participant opts in?
Anthropic says it will publish the participant’s full interview and the country associated with the conversation, while excluding account information such as name and email address.
Can a public interview be fully removed later?
Anthropic says it can remove its own copy on request, but it cannot erase copies that other people have saved.
Does the study represent public opinion about AI?
No. Anthropic explicitly says the study is not representative because all participants are Claude users, publication volunteers may differ from non-volunteers, and interview questions shape responses.
