AI-Assisted User Research: 5 Methods That Stay Rigorous

Use LLMs to speed synthesis and tagging — without replacing real users or inventing insights you never heard.

AI makes research cheaper to process and easier to misuse. These five methods keep humans in the loop while cutting hours off synthesis — the approach I use when timelines are tight on Hong Kong product work.

1. Thematic coding with LLM drafts

Feed anonymized transcripts; ask for theme candidates with quotes. You still validate themes, merge duplicates, and name insights — the model proposes, you prove.

2. Affinity digests, not affinity replacement

Cluster sticky-note text in Miro, then use AI to summarize clusters into one-sentence headers. Speeds readouts; does not remove workshop judgment.

3. Survey open-end synthesis

Hundreds of verbatims become manageable with tagged summaries — always spot-check raw responses for sarcasm, mixed language, or outliers models flatten.

4. Competitive scan with sourced search

AI search tools draft landscape tables; you verify links and pricing manually. Never ship competitor claims you did not click through.

5. Usability notes → prioritized backlog

Turn session notes into grouped issues with severity tags. Human facilitator ranks what ships; AI should not own prioritization against business context you have not given it.

Let's work together

Open to UI/UX projects, collaborations, and product design support in Hong Kong and remotely.

Let's Connect