Windtunnel Research
What AI says about a brand. What changes when the message changes. Original studies with the findings, the method, and the next question worth testing.
Subscribe to the research feed →Study
Strong sampled visibility, an inconclusive recommendation test, and six buyer profiles that turn the next content brief into a testable question.
Read the study →Study
An anonymized hotel Buyer response test: 3.45 versus 3.41, n=30 per message. Independent study; not a client.
Read the study →Study
Separate Buyer response tests, different baselines, n=5 per engine per message. Insta360 was not a client.
Read the study →Study
Five bare prompts, five repetitions, human-reviewed descriptions. Independent study; Insta360 was not a client.
Read the study →Windtunnel Research · editorial method
Windtunnel Research is the editorial byline for studies produced with Windtunnel. We verify reported figures against stored evidence, disclose methods and exact prompts, distinguish measurements from simulations, and publish negative and inconclusive findings.
Interpretations and proposed tests are labeled separately from observations. Dates identify publication and substantive updates. Existing study-specific relationship disclosures remain on their articles; publication alone implies no endorsement or client relationship.
Inspect the methodology →