How to Write a User Research Plan (2026 Template)
Qualitati Research Team · 2026-09-15 · 9 min read
Short answer
A user research plan is a one-to-two page document that ties a business decision to research questions, a method, a sample, an interview guide, an analysis plan and a review checkpoint. For AI-moderated interviews, the plan must also state what the AI moderator may probe, what it must never infer, and where a human reviews output before it reaches the roadmap.
Writing a user research plan used to be a courtesy to stakeholders. With AI moderators that can run hundreds of interviews overnight, it has become the main quality control. A weak plan no longer produces one bad study; it produces a large, confident, badly framed dataset. This guide gives you a reusable template, a scorecard to test the plan before fieldwork, and honest notes on the limits.
Key takeaways
- Start from the decision, not the method. Every research question should change what a team does next.
- In Maze's 2026 report, 69% of respondents use AI in at least some research projects, and 76% say framing the right research questions needs human involvement (Maze, March 11, 2026).
- Write the analysis plan before fieldwork, so the interview guide collects what synthesis needs.
- Give the AI moderator explicit probing rules and explicit stop rules.
- Score the plan with the 10-point readiness scorecard below before launching.
Why a user research plan matters more with AI moderation
AI moderation removes the natural brake of scheduling. When a study costs a week of calls, teams think hard before starting. When it costs a share link, they often skip the thinking. The plan puts that thinking back.
Researchers themselves draw this line. In Maze's Future of User Research Report 2026, a survey of nearly 500 researchers, designers and product professionals, respondents named interpreting nuance and emotion (82%), ethical decision-making (80%) and framing the right research questions (76%) as areas where humans remain essential (Maze, 2026). A research plan is where those three judgments are written down.
User Interviews' guide to AI moderation makes a related point: define your analysis approach before fieldwork, and watch the first few sessions for confusion (User Interviews, March 10, 2025).
The 9-block user research plan template
This is the Qualitati AI-Ready Research Plan Template. Copy the blocks into a doc; each should fit in a few sentences.
| Block | What to write | AI-moderation note |
|---|
| 1. Decision | The decision this study informs, its owner and its deadline | If no decision changes, do not run the study |
| 2. Research questions | Two to four questions the team needs answered | Humans write these; the AI never adds new ones |
| 3. Assumptions | What the team already believes, and how confident it is | List assumptions the AI must not treat as facts |
| 4. Method | Interviews, focus groups, conversational survey, or a mix | Say why AI moderation fits (or does not) |
| 5. Sample | Who qualifies, who is excluded, target number and quotas | Include a screener and fraud checks |
| 6. Guide | Topics in order, core questions, allowed probes | Set probe depth and off-limits topics |
| 7. Analysis plan | Deductive codes from research questions, room for inductive codes | Name the human who reviews AI codes and quotes |
| 8. Ethics and data | Consent, disclosure of AI moderation, data handling | Tell participants they are talking with an AI |
| 9. Review checkpoint | Pilot criteria, go/no-go, how findings are shared | No AI summary goes to the roadmap unreviewed |
Block 1–2: decision and research questions
Good research questions are answerable, bounded and tied to the decision. "What do users think of onboarding?" is a topic. "Which step in onboarding makes trial users stop before inviting a teammate, and why?" is a research question. Keep interview questions separate: research questions are for the team, interview questions are for participants.
Block 3: assumptions the AI must not infer
An AI moderator follows the guide literally and probes on what participants say. It cannot know which beliefs are hypotheses. Write them down, for example "we assume price is not the main blocker," so the guide tests them instead of leading toward them.
Block 6: probing rules and stop rules
Specify how many follow-ups per core question, which topics deserve deeper probing, and what to avoid: leading questions, personal medical or financial details, and pressure when a participant declines. See our AI-moderated discussion guide for worked examples.
Block 7: analysis plan before fieldwork
Turn each research question into a deductive code, and plan an inductive pass for surprises. Decide how you will check AI coding: sample transcripts for human review, and verify every quote used in a report against the transcript. Evidence on hard, long-form material is sobering: in one 2026 benchmark, the best LLM reached a mean F1 of only 0.41 on 567 ethnographic texts (Goodall et al., 2026). Plan review time accordingly.
AI Research Plan Readiness Scorecard
Score one point per item. Launch at 8 or more; fix the gaps first if you score lower.
- The decision, owner and deadline are named.
- Each research question maps to at least one guide section.
- Assumptions are listed and marked as untested.
- The method choice is justified against at least one alternative.
- Inclusion, exclusion and quotas are written down.
- The screener includes attention or fraud checks.
- Probe depth and off-limits topics are specified for the AI moderator.
- The analysis plan names a human reviewer for codes and quotes.
- Consent text discloses AI moderation and data handling.
- A pilot of a few sessions is scheduled with go/no-go criteria.
Worked example: a churn study plan
Decision: whether to rebuild the team-invite flow next quarter. Research questions: Why do trial accounts stop before inviting a teammate? What would make inviting feel worth it? Method: AI-moderated text interviews, because the questions are well defined and participants are spread across time zones. Sample: churned trial users from the last 60 days, excluding internal accounts. Checkpoint: pilot five sessions, read every transcript, adjust probes, then open the link. Our churn interview guide covers the fieldwork.
Where Qualitati fits
Qualitati is an AI user research platform for product, UX and customer insights teams. Once a plan is written, blocks 4 to 7 map directly onto the product: AI-moderated interviews in text and voice, AI-moderated focus groups, and conversational surveys with AI follow-ups for method; configurable interviewer instructions for the guide; and ThemeLens, which maps codes to research questions across up to 100 transcripts, with participant-anchored quotes for the analysis plan. Studies can run in 10 languages. You can start free with 30 credits, no credit card required, and check transparent pricing before planning your sample.
Limitations and methodology concerns
- Plans do not replace exploration. For truly unknown problems, open-ended human fieldwork often comes first; User Interviews also advises against AI moderation for early exploratory work.
- Over-planning can blinker. A rigid guide can miss what matters most to participants. Keep space for inductive codes.
- Scale is not saturation. More interviews do not fix a poorly framed question. See how many AI-moderated interviews you need.
- Reporting still matters. For academic work, check your plan against reporting checklists such as COREQ's 32 items (Tong et al., 2007).
Human-review note: the scorecard is an editorial framework, not a validated instrument. Adapt thresholds to your team and ethics requirements.
Who this is for, and when not to use it
This template is for product managers, UX researchers and insights teams running evaluative or well-scoped generative studies. Do not use it as-is for regulated clinical research, studies with vulnerable populations that need formal ethics approval, or ethnographic fieldwork where the question emerges on site.
FAQ
What should a user research plan include?
A decision, research questions, assumptions, method, sample, interview guide, analysis plan, ethics and data handling, and a review checkpoint.
How long should a user research plan be?
Usually one to two pages. If stakeholders will not read it, it cannot align them.
What is the difference between research questions and interview questions?
Research questions state what the team needs to learn. Interview questions are what participants are asked to help answer them.
Can AI write my research plan?
AI can draft blocks and check consistency, but a human should own the decision, research questions and assumptions.
Do I need a pilot for AI-moderated interviews?
Yes. Read the first few transcripts in full before opening the study to your whole sample.
Conclusion
A user research plan is the cheapest quality control in AI user research: nine blocks, one scorecard, and a human checkpoint before findings reach the roadmap. Write the plan, pilot it, then scale. Start free with 30 credits and run your first AI-moderated interview study on Qualitati.
Last updated: September 15, 2026. Independent editorial summary; statistics are attributed to their original sources.