The Prolific team surveyed around 3,000 participants across 15 countries about their knowledge and attitudes toward synthetic data and digital twins, finding that almost three-quarters ( 74%) expressed discomfort with the idea of an AI-generated digital twin being created from their personal data.

Almost half ( 47%) had never heard of digital twins before the survey, while 28% had never heard of synthetic data.

The survey found that 81% of participants said they would feel less valued as a participant if digital twins became commonplace, while 72% agreed with the statement that synthetic data would reduce the number of studies available to human participants.

Over nine-in-10 ( 93%) agreed with the statement that research needs real people to be trustworthy.

In a choice experiment as part of the research, half of participants ( 51%) reported that they would consider allowing a digital twin under the right circumstances, with 84% considering the ability to review and correct their digital twin as essential or very important.

A further 78% said they wanted a say in decisions about how it is used, and 82% agreed that they should own their digital twin, either independently or jointly with the organisation that created it.

Prolific has published the findings in a whitepaper, What participants think about synthetic data and digital twins.

Andrew Gordon, head of research sciences at Prolific, said: “Most of the evidence says synthetic data hasn’t reached the accuracy being claimed for it, and the only reliable way to find out what people think is still to ask them.

“Real human respondents are the foundation of this entire industry, and their perspective should count whenever we discuss methods that could materially affect their participation. This study gives them a voice.”

Methodology
The survey ran on Prolific in October 2026 with 3,046 participants, split between the UK ( 1,008 ), the US ( 1,005 ) and 13 other countries ( 1,033 ), weighted to the size of Prolific’s active pool in each. The third group came from Canada, Germany, South Africa, Italy, Spain, Poland, India, Australia, France, Brazil, Mexico, the Netherlands, and Portugal.

Participants read a short explanation of synthetic data, synthetic respondents, and digital twins, passed a comprehension check, and then gave their views. Through a choice experiment, the research also tested how participants felt about a twin built from their data, what terms they would set for one, and which terms mattered most for participants.