Home · Business · Product & Strategy · UX & prototyping

YES

As of 13 August 2026, AI can summarise your user research interviews.

Most people should hand this to a purpose-built tool.

Can you do it?

5 minutesto a draft.

30 minutesto something you’d act on.

Cost, all in£0/month

Skill neededchat-fluent

Who has to check ita colleague

What the alternative coststl;dv is a purpose-built AI meeting recorder with highlights and CRM-ready summaries.

If this goes wrong: the model turns a minority view into a general theme and your team prioritises the wrong product problem.

What to actually do

  1. Use a tool built for this

    The route this page recommends

  2. Do it yourself

    Second choice

    A chat interface, chat-fluent skill, and roughly 30 minutes until you can act on the result.

    How to actually do it

    1. Open your consent records and research plan, then confirm that the interview material can be processed by the AI service you intend to use.
    2. Gather the consented transcripts, anonymise names and unnecessary personal details, label each transcript with a participant ID, and remove sections that are outside the research purpose.
    3. Write down the research objective, participant groups, interview dates or sequence, product context and questions you asked.
    4. Paste the context and transcripts into the prompt, keeping each interview clearly separated and preserving enough surrounding text for quotes to be understood.
    5. Ask the AI to produce the summary with themes, evidence, contradictions, quotes, open questions and product implications kept in separate sections.
    6. Create a checking table with one row for every claimed theme and compare its participant IDs and quotes against the original transcripts.
    7. Mark which findings are repeated evidence, individual observations or hypotheses, then share the checked summary with a colleague before using it to change the roadmap or prototype.

    Prompt

    Summarise the user research interviews below for a product team. Use only the supplied material and invent nothing.

Research objective: [what we wanted to learn]
Participants: [brief, non-identifying description of each participant]
Interview context: [product, market, stage and relevant constraints]

Produce:
1. A short executive summary.
2. The five strongest themes, ranked by how much evidence supports them.
3. For each theme, list the participant IDs that support it and include one short exact quote where available.
4. Contradictions, outliers and views that should not be generalised.
5. User needs, pain points and workarounds, keeping observed behaviour separate from stated preference.
6. Open questions and follow-up research needed.
7. Possible product implications, clearly labelled as hypotheses rather than findings.

Do not infer demographics, motives or market size. Do not present one participant as representative of all users. Flag any claim that cannot be traced to a specific interview. Keep personal data to the minimum necessary and preserve participant IDs rather than names.

Interviews:
[Paste the consented transcripts here, labelled Interview 1, Interview 2 and so on.]

    Open it prefilled in ChatGPT or Claude, or copy it into Gemini, which takes no prefill link.

  3. Hand it to a person

    The distant third

    A person who owns the outcome does this end to end, worth it when the failure is dear.

What it gets wrong

  • AI cannot decide whether a repeated comment reflects a serious user need or merely the wording of your interview questions.
  • AI can make a minority view sound representative unless you check participant coverage and the original evidence.
  • AI cannot reliably read pauses, discomfort, rapport or the context behind an answer from text alone.
  • AI cannot decide which findings matter commercially or ethically without your product and organisational context.
  • Sending transcripts to a service can expose participant information if consent, anonymisation and access controls are inadequate.

Even on a YES, the friction has a name: judgement under ambiguity, consent and privacy and stakes of error.

How we scored this

Five axes, each scored nought to two by hand: ten means AI carries the task cleanly, and the thresholds that turn a total into YES, PARTLY or NO are published in the methodology. Each axis name links to its definition.

AxisScore (0–2)
Output2
Inputs2
Verification1
Liability1
Effort delta2
Total8 / 10

The methodology and its thresholds are published in full.

FAQ

Can ChatGPT summarise my user interviews?
Yes. Give it clearly labelled, consented transcripts and the research objective, then ask it to separate evidence, themes, contradictions and product hypotheses. Check every important theme against the original interviews before acting on it.
Can AI identify themes in user research?
Yes, it can group recurring statements and suggest themes quickly. It cannot decide whether a theme is meaningful, representative or caused by your questioning, so a researcher still needs to judge the synthesis.
Is it safe to upload user interview transcripts to AI?
Only if your consent, privacy and organisational rules allow that service to process the material. Anonymise names and unnecessary personal details, minimise what you upload and check how the service handles submitted data.
How do I check an AI summary of user research?
Trace every important claim to the participant IDs and exact transcript passages behind it. Separate repeated findings from outliers and hypotheses, then have a colleague challenge the interpretation before it informs a product decision.

Nearby answers

All UX & prototyping rulings

Assessed by gpt-5.6-luna (gpt-5.6-luna) on 2026-08-13, second-checked by an independent model. Wrong somewhere? Email [email protected] and it gets re-checked.

The newsletter

AI news, new answers and product picks, straight to your inbox.