How to Create a User Interview Guide with AI (2026): Questions Before the Call

A user interview guide written with AI is the question list for a decision you already named: who you will talk to, what you will not ask, the tasks you will watch, and how you will know you are done. The model can draft questions from that brief. It cannot invent a persona, a quote, or a finding. Those do not exist yet. The notes after the call are a different workflow.

That later workflow is how to use AI for user interview notes. The team shelf is best AI tools for UX researchers. The product decision the guide should serve is how to write a PRD with AI.

Prices are US list prices excluding tax from directory pages checked October 16, 2026. Published October 24, 2026. We did not score interview quality.

What the guide contains

Section	Who owns it	The model may
Decision	You, one sentence	Repeat it. It may not widen the study
Who	The recruiting rule you can actually run	Draft a screener from that rule. It may not add a demographic you did not name
Tasks	The product path a person verified	List steps. It may not add a feature
Questions	Open questions tied to the decision	Draft them. It may not add leading praise
Do not ask	Legal, privacy, and the topics you are not studying	Repeat the list
Done when	A number of sessions or a stop rule you set	Not declare the research finished

flowchart TD
  decision["You write the decision and the do-not-ask list"] --> draft["Claude or ChatGPT drafts the guide"]
  draft --> cut["You remove leading questions and any persona that was not in the brief"]
  cut --> wiki["Notion holds the guide"]
  wiki --> call["You run the interview"]
  call --> notes["Notes are a separate step, after consent"]

A question that asks the user to confirm your launch is a pitch. Cut it. A fake quote in a readout later is a hallucination. The check is how to fact-check AI writing.

Where the draft is written

The long brief. Claude Pro is $20 a month, or $17 billed annually. Put the decision and the do-not-ask list in a Project so session two uses the same guide. The container is Claude Projects vs ChatGPT Projects.

A second drafting seat. ChatGPT Plus is $20. Business is $20 a user a month billed annually, minimum two, when the brief includes unreleased product detail. The ladder is ChatGPT Go vs Plus vs Pro.

The record the team opens. Notion AI Business is $20 a seat a month. The guide, the screener, and the session log live together. Plus at $10 is not the full AI tier. The invoice is Notion AI pricing explained.

Notes, only after the guide is done and the participant agreed. Granola Business is $14 a user a month and does not join as a bot. Basic is free with 30 days of history. Fathom can join as a bot. Premium is $16 a month billed annually. Do not send a bot because the guide was easy to write. Consent comes first. The pair is Granola vs Fathom.

A prompt that holds the line:

“Draft an interview guide for the decision in my notes. Include who we will recruit, tasks, open questions, and the do-not-ask list I pasted. Do not add personas, sample quotes, or findings. Flag any question that leads the participant. If the decision is missing, stop and say Missing.”

The person who runs the session is what human-in-the-loop AI is.

What this guide is not

It is not a survey, and it is not a training quiz. Quizzes are how to create training quizzes with AI. It is not a sales demo script. That workflow is how to create a demo script with AI. It is not the readout. You cannot summarize findings before the sessions happen.

FAQ

How many questions is enough?

Enough to make the decision, few enough that a person can talk. If the model returns thirty questions, group them under the tasks and cut until the session fits the time you booked.

Can AI write the screener?

Yes, from the recruiting rule you paste. No, from “find me users who love us.” A screener that seeks praise will fill the calendar with the wrong people.

Should I paste old interview transcripts into the guide prompt?

Only as background you label as past research, and only if you are allowed to. The new guide should not treat old quotes as this study’s result.

Can the guide update itself after each call?

The questions stay stable unless you write down why you changed them. A guide that mutates every session is not a study. Put changes in the log with a date.

Last verified

October 16, 2026 for Claude, ChatGPT, Notion, Granola, and Fathom. Published October 24, 2026.

Sources: Claude pricing, Notion pricing, Granola pricing.

Published by Andrew for AI Tools Optimizer.

Notion AI is in the productivity and notes directory. Directory pages may use affiliate links. A purchase through them may earn us a commission at no extra cost to you.