As of 13 August 2026, AI can analyse usability test recordings.

What to actually do

1. Hand it to a person

   The route this page recommends

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

2. Use a tool built for this

   Second choice

3. Do it yourself

   The distant third

   A chat interface, chat-fluent skill, and roughly 1 hour until you can act on the result.

   How to actually do it

   1. Open the research plan, task script and prototype notes, then write the product, objectives and participant context into the prompt without including names or other identifying details.
   2. Gather the recording files and any existing transcript, check that the participants gave permission for this analysis, and remove files you are not authorised to process.
   3. Open an AI chat or recording-analysis tool, attach the permitted files, paste the prompt and add the research context before asking for the analysis.
   4. Export or copy the report, then open each cited timestamp in the original recording and compare the claimed behaviour and quotation with what actually happened.
   5. Mark any unsupported interpretation, invented quotation or missing uncertainty in the report, and ask the AI to correct only those specific rows without changing the evidence.
   6. Share the checked findings with a UX or product colleague, agree which issues need further testing, and keep the recording and evidence table with the final research record.

   Prompt

   Analyse the attached usability test recording and any transcript as a UX research assistant. Context: [PRODUCT OR PROTOTYPE]. Research objectives: [OBJECTIVES]. Tasks given to the participant: [TASK SCRIPT]. Participant description, using only non-identifying information: [PARTICIPANT CONTEXT].
   
   Produce:
   1. A short session summary.
   2. A table of observed behaviours, with timestamp, exact quotation or precise description, task, and whether the point is directly observed or inferred.
   3. Usability issues grouped by theme, with supporting timestamps and evidence.
   4. For each issue, state the likely user impact, affected task, strength of evidence, and any competing explanation.
   5. A separate list of positive findings and unmet needs.
   6. Recommendations ranked by evidence and potential impact, clearly labelled as recommendations rather than observations.
   7. Questions that need further research.
   
   Do not invent quotations, events, causes, severity ratings or participant characteristics. Do not treat the participant's opinion as proof of a usability problem. Distinguish what happened on screen from what the participant said and from your interpretation. If the recording is unclear, say so and give the timestamp. Keep the analysis tied to the stated research objectives. Do not include names, faces, contact details or other identifying information in the report.

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