How to do user research with AI (without a UX team)

Doing user research with AI is no longer just for big teams with a UX budget. If you have a product, a service or even an idea, you can truly understand your users using an AI like Claude or ChatGPT as your research assistant: it helps you prepare interviews, design surveys, analyze feedback and synthesize findings into something actionable.

Keep the key thing in mind: AI does not replace talking to real people. The data comes from your users; AI helps you get the most out of it. Tap each step to open and follow it.

Before you start: be clear on what you want to learn. “Why do people abandon signup?” is a better research question than “do they like my app?”. AI helps you sharpen this in Step 1.

1 Define what you want to learn

Every good research starts with a clear question. Open Claude or ChatGPT and ask it to help you focus it.

Example prompt:

I'm a [your role] and I have [product/service]. I want to
understand [the problem, e.g. why new users don't come back
after day one]. Help me turn this into 3 clear research
questions and tell me which method (interviews or survey)
fits each one.

The AI returns sharper questions and tells you whether an interview (to understand the “why”) or a survey (to measure “how many”) fits better.

Simple rule: interviews for depth, surveys for volume.

2 Prepare your interview script

Ask the AI for an interview script based on your question. A good script avoids questions that “push” the answer.

Create a 20-minute interview script to talk with [your user
type] about [topic]. Include an opening ice-breaker, 5 to 7
open, non-leading questions, and a closing one. Avoid yes/no
questions.

Review the script yourself: does any question already assume the answer? Ask the AI to rewrite it more neutral.

Tip: also ask for follow-up questions (“can you tell me more about that?”) so you’re never stuck mid-interview.

3 Design a survey people actually finish

If you need volume, build a short survey. AI is great at drafting clear questions in the right order.

Design a survey of at most 8 questions to measure [what you
want to know]. Mix multiple-choice and one or two open
questions. Order them from easy to personal. Also write a
short intro explaining what it's about.

Move them into a free tool like Google Forms or Tally and share it. To go deeper on this step, see our guide on creating forms and surveys with AI.

4 Analyze the feedback with AI

This is where AI saves you hours. Instead of reading 40 responses by hand, you hand them over and ask it to find the patterns.

First, clean the data: remove names, emails and anything that identifies a person.

Then:

Here are [30 responses / the transcript of 6 interviews].
Find the 5 most repeated themes, give me one representative
quote for each, and flag any complaint or need that shows
up more than once.

The AI returns the grouped themes. You read them with your judgment: does any surprise you? does any confirm what you already suspected?

Don’t upload personal data without checking the tool’s privacy policy.

5 Synthesize findings into something actionable

Themes alone are useless if they don’t turn into decisions. Ask the AI to build the synthesis.

With these themes we found, write a findings summary with:
the 3 main user problems, how serious each one is, and one
concrete recommendation per problem. Clear format to present
to my team.

You can also ask it to build a simple user profile (persona) from what you learned, or a prioritized list of what to fix first.

You make the final call: AI organizes and proposes, but you know your product and context better than any machine.

If something goes wrong

Problem	Fix
The AI invents data or users	Give it only your real responses and say “use only the information I gave you”
Themes come out generic	Ask for verbatim quotes from the responses to anchor each theme
I have few responses	Start with 5-8 interviews; patterns show up sooner than you think
I’m worried about privacy	Anonymize everything before uploading and check the tool’s policy

When you want to use this to study your competition, continue with how to research your competition with AI.

Frequently asked questions

Can AI replace talking to real users?

No. AI helps you prepare, organize and analyze the research, but the data has to come from real people. AI doesn't invent your users; it helps you understand them better.

Do I need UX knowledge to do this?

No. With this guide and an AI like Claude or ChatGPT you can prepare solid interviews and surveys from scratch. The AI suggests the questions and teaches you the method as you go.

Is it safe to upload my customers' feedback to an AI?

Remove names, emails and identifying details before uploading anything, and check the tool's privacy policy. Use AI to analyze patterns, not to store personal data.

How many responses do I need to draw conclusions?

For qualitative patterns, 5 to 8 solid interviews already start showing repeating themes. For number-based surveys, the more responses the more confidence; but even 20 or 30 give you useful early clues.

Which AI should I use, Claude or ChatGPT?

Either works. Claude tends to handle long transcripts and lengthy text better; ChatGPT is handy for quick brainstorming. Try both and keep the one that fits you.