Buy survey and market research data for AI training
Last updated How we research
In short
Survey and market research data is respondents' answers, including free-text responses and repeat panel waves, licensed to train and evaluate models on how people think and choose. Consent is the main constraint: the Insights Association Code says personal data should not be used in AI training data sets without informed consent. New fieldwork has published costs, with clickworker rewards from $0.25 per participant plus a 40% fee.
Key takeaways
- AI teams use survey and market research data for (1) evaluating whether model answers match real opinion distributions, (2) training models to code and summarize open-ended answers and (3) simulated respondents and demand forecasting.
- Usually priced per completed response for new fieldwork, per study for existing archives.
- Common delivery formats: Respondent-level tables in CSV, SPSS SAV or Parquet, codebooks and questionnaire files and open-ended answers in JSONL with question IDs.
- Check before you license survey and market research data: (1) respondent consent covered AI training, as the Insights Association Code requires, (2) ICC/ESOMAR Code: separate consent for non-research use and (3) US: CCPA, including sensitive answers such as health, religion or sexual orientation.
- Every seller we shortlist passes 8 vetting checks before you see a sample.
Survey and market research data: buyer's spec
- Used for
- Evaluating whether model answers match real opinion distributions
- Training models to code and summarize open-ended answers
- Simulated respondents and demand forecasting
- Preference and behavior modeling by segment
- Delivery formats
- Respondent-level tables in CSV, SPSS SAV or Parquet
- Codebooks and questionnaire files
- Open-ended answers in JSONL with question IDs
- Weights and panel wave identifiers
- Usually priced
- Per completed response for new fieldwork, per study for existing archives
- Check before you sign
- Respondent consent covered AI training, as the Insights Association Code requires
- ICC/ESOMAR Code: separate consent for non-research use
- US: CCPA, including sensitive answers such as health, religion or sexual orientation
- US: COPPA and verifiable parental consent for respondents under 13
- US: HIPAA if a covered entity ran a health survey
- Client ownership of commissioned studies
- Open-ended answers scrubbed of names and contact details
- UK and EU: UK GDPR, EU GDPR and the MRS Code of Conduct
Why AI teams work with us
- 1
Every seller of survey and market research data we shortlist has passed eight vetting checks, from authority to sign through to a re-identification test.
- 2
Each shortlist comes with a vetting file per seller: rights evidence, provenance record, de-identification method, data manifest and proposed license terms.
- 3
Samples are drawn at random across the archive and labeled with the sampling method, so you can judge the whole package rather than its best rows.
- 4
You contract directly with the seller, and we help agree acceptance tests, staged payments and remedies before any data is delivered.
What is survey and market research data?
Survey and market research data is the respondent-level record of studies: closed answers, ratings, open-ended text, demographics and, for panels, the same people's answers over time. It comes with a questionnaire and a codebook that explain each variable.
AI buyers value it because it links stated views to the kind of person who holds them, which web text does not. Open-ended answers add natural language written in response to a known question.
What do AI teams use survey data for?
AI teams use survey data mostly for evaluation and for modeling people. GlobalOpinionQA, published by Anthropic, is built from questions and answers in cross-national surveys and was designed to measure how well model answers represent opinions in different countries. Its Hugging Face card lists a CC BY-NC-SA 4.0 license.
- Evaluation: comparing model answers with real answer distributions by country or segment
- Open-end coding: training classifiers and summarizers on human-coded verbatims
- Simulated respondents: models that predict how a segment would answer a new question
- Forecasting: linking stated intent to later panel behavior
Our page on evaluation datasets covers what makes a good test set.
Do respondents have to consent to AI training?
Respondents have to consent to AI training under the main industry codes, even where US law is less specific. The Insights Association Code, last updated September 2026, says personal data should not be used in AI training data sets without informed consent, and requires respondents' consent before personal data passes to third parties unless the law allows otherwise.
The ICC/ESOMAR International Code says separate consent must be obtained for non-research purposes, and limits personal data to the purpose for which it was collected. Ask to see the consent wording respondents accepted, and check that it names AI training or licensing to third parties.
If a panel changes its terms to allow training, note that the FTC has warned that quietly changing terms of service could be unfair or deceptive. Older studies collected under narrower terms are the riskiest part of any archive.
Which US privacy laws apply to survey data?
The US privacy laws that apply to survey data depend on who answered and what they were asked.
| Rule | When it applies | What to check |
|---|---|---|
| CCPA | Responses from California residents held by a covered business | Sensitive answers such as health, religion or sexual orientation |
| COPPA | Online surveys collecting personal information from children under 13 | Verifiable parental consent and separate consent for AI training disclosures |
| HIPAA | Health surveys run by or for a covered entity | De-identification by Safe Harbor or expert determination |
| Industry codes | Members of the Insights Association or ESOMAR | Informed consent for AI training and third-party transfer |
Panel data raises re-identification risk, because many waves of answers about one person can single them out without a name. The Insights Association Code also asks researchers to ensure personal information cannot be reverse-engineered by AI inference. Our de-identification guide covers the methods.
How much does survey data cost?
Survey data has published costs only for new fieldwork. No buyer or marketplace we checked publishes rates for existing archives, and Datarade lists panel survey datasets with pricing available on request.
- Prolific: participant pay of at least $8.00 an hour, $12.00 recommended, plus a 42.8% platform fee for corporate customers
- clickworker, part of LXT: rewards from $0.25 per participant plus a 40% fee, according to LXT's June 2026 announcement
At Prolific's recommended rate, a 10-minute survey costs $2.00 per respondent before the fee and about $2.86 with it, on our arithmetic from its published pricing. See the Prolific and LXT profiles.
Who owns a market research archive?
A market research archive can belong to the clients who commissioned each study rather than the agency that ran it, depending on the contract. Check each client agreement before licensing an agency's back catalog.
Open-ended answers are text written by respondents, so check the panel terms for a license from respondents as well as their consent.
If you hold UK or EU data
If respondents are in the UK or EU, their answers are personal data under UK GDPR or EU GDPR until anonymized, and answers about health, religion or politics are special category data. UK researchers also work to the Market Research Society's MRS Code of Conduct, whose current edition took effect in May 2023.
How do you source vetted survey data?
You source vetted survey data by naming the countries, topics and panel waves you need, then checking the consent wording behind each study. Every seller we shortlist passes our eight-point vetting standard. Book a free consultation or submit a data requirement.
Frequently asked questions
Can a market research agency sell survey data for AI training?
Are open-ended survey answers personal data?
How much does it cost to commission survey data for AI?
Does the ESOMAR code apply to US research firms?
Can synthetic respondents replace real survey data?
Sources
- Code of Standards, Insights Association, accessed November 2026
- ICC/ESOMAR International Code on Market, Opinion and Social Research and Data Analytics, ESOMAR and International Chamber of Commerce, accessed November 2026
- Towards Measuring the Representation of Subjective Global Opinions in Language Models, arXiv (Durmus et al., 2023), accessed November 2026
- Anthropic/llm_global_opinions dataset card, Hugging Face, accessed November 2026
- Pricing, Prolific, accessed November 2026
- LXT launches fully integrated self-service survey platform through clickworker, LXT via PR Newswire, accessed November 2026
- Survey data, Datarade, accessed November 2026
- California Civil Code section 1798.140 (CCPA definitions), California Legislative Information, accessed November 2026
- Children's Online Privacy Protection Rule (final rule amendments, 90 FR 16918), Federal Register, accessed November 2026
- Guidance regarding methods for de-identification of protected health information, US Department of Health and Human Services, accessed November 2026
- AI (and other) Companies: Quietly Changing Your Terms of Service Could Be Unfair or Deceptive, Federal Trade Commission, accessed November 2026
- MRS Code of Conduct, Market Research Society, accessed November 2026
For AI teams
Ready to source?
Book a consultation to brief us. We shortlist vetted sellers, each with a vetting file covering rights evidence, provenance and a representative sample.
Related guides
- De-identification for AI training data
Legal and compliance
- How to buy AI training data
Buying data
- AI training data due diligence checklist
Buying data
- Sourcing expert and human-generated data: vendors, costs and quality
Buying data
- Buying evaluation datasets: what makes a good AI eval set
Buying data