UX research produces a lot of raw material: interview recordings, notes, survey answers, support tickets and usability sessions. Turning that into findings is slow, and that is where AI helps most.
Common uses today include automatic transcription, summaries of individual sessions, clustering notes and open survey answers into themes, and searching across many studies at once. AI is also useful for preparation: drafting interview guides, screener questions and survey wording for a researcher to review.
The risky use is replacing participants. Some tools offer "synthetic users", AI personas that answer research questions. They can help you prepare, but they repeat what is common in their training data and cannot tell you what your specific users actually do. Treat them as a brainstorming aid, not evidence.
Summaries can also be wrong. An AI may overstate a theme, miss a rare but important problem, or invent a quote. Good practice is to link every finding back to real quotes or clips, and to read a sample of the raw data yourself.
Privacy matters too. Remove personal details before sending recordings or transcripts to an AI service, and check that consent covers it. Used carefully, AI frees researchers to spend more time with users, which is the part that matters most.