Claude interviewed almost 81,000 people for Anthropic

Almost 81,000 people in 159 countries, writing in 70 languages, sat down for an interview about AI, and Claude ran every one of those interviews. Anthropic describes the setup on its Anthropic Interviewer page: a research tool built on Claude that runs detailed conversations at scale and hands the results back to researchers.

At a glance

• Anthropic Interviewer is a research tool built on Claude that plans, conducts and helps analyse conversational interviews, with Anthropic's researchers setting goals at the start and checking findings at the end.
• Besides the global pilot of almost 81,000 participants, Anthropic used the same tool for a focused test study on how workers, scientists and creatives fit AI into their professional lives.
• Claude drafts the interview guide, asks the questions and summarises the answers, which leaves much of the weight on the researchers who refine the plan and validate the findings.

If you haven't been following Anthropic's research, its earlier work showed a lot about what people use Claude for. It said far less about how Claude's outputs affect people, how they feel about AI and what role they picture it playing in their futures. Anthropic says the only way to get that picture, and to keep people at the centre of how the technology gets built, is to ask them directly.

The headline study is a global research pilot. Almost 81,000 people across 159 countries and 70 languages described in their own words how AI has affected them and what they want from it. Anthropic calls it a pilot. It ran alongside the focused test study of workers, scientists and creatives.

Anthropic says the findings feed its Societal Impacts research and help shape how it builds Claude. Under its policy commitments, the company shares what it learns with policymakers, independent researchers and the public. Before agreeing to an interview, each participant is told how their responses will be used, and the tool has its own privacy policy.

The tool works in three stages. In planning, researchers set the study's questions and goals. Claude then drafts an interview guide, meaning a strategy and a list of questions based on established practice, which the team refines and signs off. In the interview itself, Claude talks with the participant in real time, in that person's own language, and picks each follow-up based on what they just said.

Think of a printed survey compared with a good reporter holding a notebook. The form asks everyone the same thing. The reporter hears an answer and asks about the interesting part. In the analysis stage, the tool checks the responses against the original research plan, groups them into themes as they emerge and pulls out quotes that illustrate them, while researchers work alongside it to validate and interpret the results.

The page does not explain the method behind the numbers. It does not say how many researchers checked the work, how themes were validated across 70 languages, or how answers from 159 countries were balanced against each other. In our view, the validation step carries most of the weight, because Claude both asks the questions and summarises the answers, so a blind spot in the interviewing may carry straight into the analysis.

Where the 81,000 answers go

Anthropic points readers to the surveys it has already published from Anthropic Interviewer. The page gives no date or scope for the next study. It also doesn't say whether outside researchers will get the tool itself or only its findings. Until that changes, policymakers and the public will see these conversations through Anthropic's own write-ups.