Help me design a user-research study that can inform one specific product decision. Treat AI as a planning assistant, not as a source of user evidence. Product and user context: [PRODUCT AND USERS] Decision the research must inform: [DECISION] What we already know, including evidence and assumptions: [CURRENT EVIDENCE] Study constraints: [CONSTRAINTS] Research method I am considering, if any: [PREFERRED METHOD] Ask up to 3 clarifying questions only if a required input is missing. Then produce a research plan with these sections: 1. Decision statement, primary research question, and 3-5 subquestions. Separate descriptive, evaluative, and generative questions. 2. Participant criteria, exclusion criteria, target sample, recruitment screener questions, and rationale. Do not claim a qualitative sample represents a percentage of all users. 3. Recommended method and session structure. If interviews, give a 45-minute discussion guide with neutral, open questions, follow-up probes, and timing. If usability testing, specify realistic tasks, success signals, and moderator prompts that do not lead participants. 4. A note-taking framework, analysis plan, and evidence table with columns for observation, source, confidence, alternative explanation, and implication. 5. Risks, accessibility considerations, consent/privacy issues, and what this study cannot answer. Avoid leading questions, solution-first wording, hypothetical “would you use” questions, and claims that need behavioral data we do not have. Do not invent participants, quotes, findings, or market facts. Before answering, self-check that every question links to the decision and that the study could disconfirm the team’s current assumption.
Fill in
| Placeholder | What to enter | Example |
|---|---|---|
| [PRODUCT AND USERS] | Describe the product, target users, their context, and the part of the experience you are studying. | Mobile expense-management app for self-employed contractors; studying receipt capture after a job site visit. |
| [DECISION] | State the specific decision the research should help make. | Decide whether to build automatic receipt reminders tied to bank transactions or improve the manual camera flow first. |
| [CURRENT EVIDENCE] | List existing research, product data, support themes, and assumptions separately. | Eight support tickets mention lost receipts; analytics show 42% abandon camera capture; assumption: users forget until tax time. |
| [CONSTRAINTS] | State timeline, budget, participant access, privacy limits, and any research operations constraints. | Two weeks, no incentive budget, access to 12 active users, cannot record bank-account details. |
| [PREFERRED METHOD] | Name a method you are considering or write that you want a recommendation. | Recommend the best method, likely remote interviews with a short workflow walkthrough. |
How to use
- Fill in the decision before describing possible features, so the study is not framed to justify a solution.
- Paste evidence with its source and date, keeping assumptions in a separate line.
- Check that screener criteria match the user segment that will experience the decision.
- Follow up with: “Rewrite the interview guide for participants who have abandoned receipt capture at least twice, without revealing our reminder idea.”
Variations
Interview guide
Use this when the study plan is settled and you need a moderator guide.
Create a 45-minute semi-structured interview guide for [PARTICIPANT TYPE] about [BEHAVIOR OR PROBLEM]. The decision it informs is [DECISION], and known context is [EVIDENCE]. Include a warm-up, chronological walkthrough of a recent real event, neutral probes, artifact prompts, and closing questions. Mark questions to avoid because they are leading or hypothetical. Do not mention our proposed solution, [PROPOSED SOLUTION], unless it is reserved for a final concept test. Include time boxes and a note-taking template.
Usability test
Use this to test a prototype or live flow with realistic tasks.
Design a moderated usability test for [PROTOTYPE OR FLOW] with [PARTICIPANT TYPE]. Our learning goals are [LEARNING GOALS], and the realistic scenario is [SCENARIO]. Produce a 30-minute script with introduction, consent reminder, task prompts, observation points, rescue rules, and closing questions. Define success signals without turning them into pass-fail judgments about participants. Do not lead users toward interface elements or state that the design should be easy. Flag accessibility and device considerations.
Findings report
Use this after research is complete and notes need disciplined synthesis.
Analyze these user-research notes: [NOTES]. The research questions were [RESEARCH QUESTIONS], and the participant details are [PARTICIPANT DETAILS]. Produce a decision-ready findings report with themes, source-linked observations, disconfirming evidence, confidence notes, and implications. Keep direct quotes separate from interpretation. Do not assign population percentages or causal claims from this qualitative sample. End with opportunities ranked by evidence strength, plus unanswered questions and the next validation step.
Tips
- Recruit for the behavior under study, not just a broad demographic; someone who recently completed the workflow can recall real workarounds and triggers.
- Ask participants to reconstruct a recent event step by step before showing a concept, because recalled behavior is stronger evidence than preference statements.
- Preserve exceptions in your synthesis; a finding that applies only to new users or one operating context should not become a universal requirement.
- If research includes sensitive health, financial, employment, or account data, minimize what you collect and use your organization’s approved consent and storage process.
FAQ
Can ChatGPT conduct user research for me?
It can help plan, transcribe, structure notes, and critique study materials. It cannot substitute for talking to representative users or observing their real behavior.
How many participants should I recruit?
It depends on the method and decision. Start with enough participants to cover key segments and watch for recurring patterns, while avoiding statistical claims from a small qualitative sample.
Should I share raw transcripts with AI?
Only if your privacy policy, consent language, and approved AI tools allow it. Remove unnecessary personal or sensitive information first.