SN Research · Survey data integrity for market and social research
Can you trust your survey responses to be human and not AI?
This product is seeking 3 Foundational Software Partnerships to grow the product.
Find out moreThe problem
Survey research rests on one assumption: that the people in your data are real, and that they answered genuinely. That assumption is failing.
In August 2026, Pew Research Center tested 11,114 respondents from online opt-in sources and found that 18% failed trap questions designed to catch people who are not answering genuinely. A study published in PNAS in November 2025 showed an AI respondent passing 99.8% of attention checks, at around five cents a response. Click farms run at volume, and the fraud-screening tools disagree with one another: in one peer-reviewed comparison, the choice of tool alone moved a study's results by 8.5 percentage points.
The cost is not a line in a quality report: it's a finding published, a product launched or a policy funded on the strength of people who were never there. Excluding a genuine respondent loses one data point; including a fraudulent one can change what you report as true.
What is SN Research?
SN Research is patent-pending software, designed and built in Australia by 01Mind for organisations whose work depends on survey data. SN Research runs on your own server, examines every respondent in data you have already collected, and gives you back a qualified dataset: every respondent scored out of 100 (the lower the score, the more likely the respondent is not who they claim to be), every concern explained, every finding graded by confidence, and a finished report you can put in front of a client, an ethics committee or a journal.
What SN Research does
The software uses 25 vectors that cross-check one another to test for four kinds of threat, organised under the Total Survey Error framework that survey statisticians already use:
- AI-generated and fabricated responses: answers a machine wrote, answers disguised to hide that and answers a respondent did not genuinely give.
- Sampling, coverage, non-response and weighting: whether your sample and your weighting can support what you report.
- Organised incentive fraud: respondents sharing devices, locations or timing, and professional respondents answering at volume.
- Translation and localisation: whether a translated instrument still asks the same question.
Every check produces a graduated finding with the evidence behind it, never a bare accusation. Your researchers make the decision; SN Research makes sure that decision rests on what was actually measured.
Built for how research works
- Many independent signals, not one detector. Every serious threat is examined by several checks that look at different properties of the data, so a respondent working to get past one makes themselves more visible to another.
- Corroboration before consequence. No respondent is excluded on a single signal. A single signal only raises a flag when it is overwhelming, and a flag only asks a researcher to look, which protects the genuine respondent who writes simply, answers quickly or holds a mixed view.
- Your instrument, not only your respondents. When many respondents stumble over the same question, SN Research names the question as the likely cause, rather than blaming everyone who answered it.
- Your data never leaves your server. All analysis runs on your own hardware, including the two AI models the software uses. It never keeps survey responses, names or contact details, only a respondent ID and the scores it calculated, and it is built so that none of your data or reports is ever sent to us. The only thing an installation sends 01Mind is a diagnostic licence check-in, and the Software Licence Agreement lists exactly what it may contain: never your data, and never your reports.
- The software works with the files you already export. CSV, TSV and Excel workbooks, including Qualtrics and SurveyMonkey exports, with no cap on the number of respondents in a file.
- Fast enough for fieldwork. Measured on a developer machine, 1,000 respondents with ten open-ended questions each are scored in about two minutes.
- Reports you can use. Every upload produces a Word report written to professional reporting conventions, and an Excel workbook of scores. Both can go straight into a researcher's final report.
- The software learns from your researchers. When a researcher confirms what a respondent really was, SN Research adjusts how much each check counts, within fixed limits and only once there is enough evidence. What it learns stays on your server and is tuned uniquely to how you use it.
- Simple to buy. No subscriptions. An annual licence in two tiers, with any number of users.
Pricing
Who it is for
Market and social research companies, panel providers, and government and academic research units that collect or depend on survey data.
Try SN Research free
Your organisation can trial SN Research free for 7 business days, installed on your own server and used on your own data. You can put as much data through it as you want. We offer one trial per organisation.
At the end of the trial, you have two business days to tell us whether to continue. If you say yes, you choose your tier, your licence starts that day, there is nothing to reinstall, and we invoice on 30-day terms. If you say no, or do not answer, the software removes itself automatically from your system. It takes nothing it didn't come with.
Take a FREE 7-business-day trial
This product is seeking 3 Foundational Software Partnerships to grow the product.
Find out more