AI tools can make user research faster to plan and easier to organize. They can also tempt a team to skip the slow part: talking to the people they are designing for. The most useful approach treats AI as a research assistant that helps you listen better, not as a stand-in for the people you need to hear from.
Map what you already know
Many teams have more evidence than they realize: support tickets, sales call notes, app reviews, survey comments, and search terms from the help center. With permission and with identifying details removed, an AI assistant can help you group that material, list recurring complaints, and highlight where customers describe the same problem in different words.
Ask it to separate what customers said from what it is inferring, and to quote the source text for each pattern. Then open a sample of the originals yourself. The goal is a sharper list of questions, not a finished conclusion.
Prepare for better conversations
AI is useful before an interview. Give it your research goal and ask for neutral, open questions about past behavior. Ask it to flag any question that is leading, hypothetical, or assumes a solution. Ask what you might be missing about people with different access needs, devices, languages, or levels of experience.
You can also ask it to play a skeptical colleague: “What would make these findings unreliable?” That kind of critique strengthens the plan before you spend a participant’s time.
Be careful with synthetic users
Generated personas and simulated interviews can sound remarkably realistic. They are built from patterns in text, not from your customers’ circumstances, so they tend to produce plausible, average answers and miss the surprising details that make research valuable. Use them to rehearse your interview guide or to brainstorm questions. Do not present their answers as findings about real people.
Turn observations into shared understanding
After sessions, AI can help draft tidy notes, suggest candidate themes with supporting quotes, and produce a first version of a journey map. Keep every claim traceable to a session and a timestamp. When a theme rests on one vivid participant, say so. When people disagreed, show the disagreement instead of smoothing it away.
Keep empathy human
Understanding someone includes noticing tone, hesitation, and context that never appears in a transcript. Watch at least some sessions live. Share short, consented clips with your team. The point of user research is not a document; it is a team that understands the people it serves well enough to make better decisions.
Try it yourself
Gather twenty anonymized support messages or reviews for one product. Ask an AI assistant to group them and quote evidence for each group. Then read all twenty yourself and write down one pattern the AI missed and one it overstated.