Synthetic survey data
Realistic respondents for researchers, analysts and students.
Paste your questions, say who should answer, and download a respondent-level data file in 5–20 minutes. Every respondent is based on a real person's record from the US Census. Their answers are generated.
$49 per data file
up to 120 questions · 5,000 people · no subscription
anonymous Census records to draw your respondents from, in the right proportions.
real respondents: what a file from your questions alone was worth in our tests. See the results.
per data file, up to 120 questions and 5,000 people, ready in 5–20 minutes. No subscription.
Grounded in real Census records
Each respondent starts from an anonymous American Community Survey record (2020–2024, 16 million of them), drawn in proportion to the population you choose. Age, sex, education, state and employment come from that record. Everything else, including every answer to your questions, is generated: the respondents are synthetic, not real people.
A data file, not a chatbot
One row per respondent, one column per question, ready for SPSS, R, Stata or Excel, with a codebook and a read-me that says what the file is worth.
Good for
- Testing a questionnaire before you pay for fieldwork
- Designing and testing a study before IRB review or fieldwork
- Finding weak, redundant or confusing items
- Planning sample sizes and analysis code on realistic data
- Teaching survey methods and statistics
Not for
- Replacing real respondents in a published result
- Decisions about individual people (hiring, credit, care)
- Very narrow groups, or populations outside the US
- Significance tests that treat the rows as real people
How it works
- Your questions. Rating scales, select-all-that-apply, numbers and Census demographics, mixed in one survey.
- Who answers. Age, sex, education, employment and states, plus a plain-English description such as “who work in healthcare”.
- Generate. An AI model reads each question and estimates how this population answers; a statistical model turns that into individual people with consistent answers.
- Download. CSV, codebook and a read-me that says what the file is worth.
Questions? Read the FAQ.