1
Share the study setup.
We agree on questions, audience, and scoring.
Blind validation · The Lewsearch Blind Challenge
You have the answers. We don't.
Bring a completed consumer study. We'll run the agreed questions with simulated Americans before seeing your results. You decide how we measure up.
Your first qualified U.S. consumer Blind Challenge is free.
Keep your human results. No purchase required.
Put Lewsearch to the test.
Leave your details. We'll confirm fit and arrange the questionnaire separately. Please keep your human results with you.
2
Receive our predictions.
We deliver the readout before seeing your results.
3
Compare the results.
You score the agreed questions against your benchmark.
Public score
On the calibrated 50-state benchmark, the average miss is 7.07 points on the best 80% of 1,521 scored questions, and 10.02 points across all of them.
See the evidence →Questions
Is the Blind Challenge free?
Yes. Your first qualified Blind Challenge study is free. There's no obligation to purchase anything afterward.
Do I need to share the human results?
No. You can compare privately after receiving our readout. If you want to discuss the comparison, you can share summary feedback afterward; respondent-level human data is not required.
What do you need to run the study?
After confirming fit, we'll ask for the agreed question wording, answer choices, target-audience criteria, original field dates, and relevant study-design details. Answer choices are the options on the questionnaire. Human answers are the results we must not receive before delivery. Do not include outcome tables, findings, or clues about which concept won.
Can I use a study whose results are already public?
Tell us if the results are public, but do not send the result link during blind setup. This is a blind comparison against the results you withhold, not a guarantee that a model has never encountered related public information. An unpublished human benchmark makes the strongest test. We'll discuss whether to proceed and how to label the comparison.
What about focus groups or open-ended studies?
Ask us about fit. Qualitative comparisons need a different evaluation approach; they are not interchangeable with topline accuracy. U.S. consumer multiple-choice studies are the starting point.
What if Lewsearch misses?
That is useful information too. Compare the agreed questions using the agreed method, including the misses. One study can help you judge fit for that use case; it does not establish universal accuracy.
Is this a replacement for human research?
No. Lewsearch uses simulated consumers, not recruited human respondents. This challenge is a way to evaluate where synthetic output may be useful in your workflow.
Will my study become a public case study?
This request does not grant permission to publish your study, results, or company name. Any case study would require separate permission.
What happens after I request a challenge?
We'll review whether your study is a qualified U.S. consumer fit and confirm scope before the run. Your first qualified Blind Challenge is free. Submitting a request does not create a purchase obligation. If you need confidentiality terms, raise that before sharing your instrument. See the Privacy Policy.