Free Tool
A UX research diagnostic for AI products that behave differently every time.
A free, fillable worksheet that surfaces — in about five minutes — where your standard methods are most likely to hand you a false signal on an AI feature. Fill it in before you write your research brief, and bring it to your study-design conversation.
Fillable PDF
2 pages · ~5 min
AI Product Research Diagnostic
Most research methods were built for products that respond the same way every time. AI products don't. This diagnostic walks you through the feature you're about to study, rates it against the five challenges that quietly break standard methods, and helps you name which one is most acute — so you walk into study design knowing where the risk actually is, not where you assumed it was.
Scroll to preview both pages · fillable PDF
The five challenges it checks
- 1Unpredictable behavior — different output for different participants, or the same participant twice.
- 2Mental model instability — participants reacting to entirely different imagined systems.
- 3Trust as behavior — what people accept, verify, delegate, or override, not what they say.
- 4Explainability demands — confidence scores without the provenance users need to judge them.
- 5The moving-target problem — a model that ships updates between your brief and your findings.
Copyright © 2025
Created with
