Moderated & unmoderated user research · PUBLIC RESEARCH BRIEF
LookbackWhat makes an AI-suggested finding worthy of the research record?
Lookback describes moderated sessions with live stakeholder observation and unmoderated studies with AI follow-ups, summaries and project-level exploration. It says AI suggestions are labeled and can link insights to moments in recordings. Traceability helps, but a timestamped excerpt still needs study context, contradictory evidence and human judgment before it becomes a finding.
Updated 2026-10-21 · Simulation results not yet generatedCHANGE ONE THING. LEARN WHAT MATTERS.
Three questions for the GTM team.
How should live observers take notes and discuss a moderated session without steering the moderator or turning the first vivid reaction into a team conclusion?
Set up this study →When unmoderated AI follow-ups produce more detail than the original task requested, which responses belong in the finding and which require a separately approved question?
Set up this study →PROPOSED AUDIENCE
Who should weigh in?
North American UX researchers, product designers, research operations, product managers and stakeholders who observe or consume Lookback studies. Include moderators, unmoderated-study owners, observers and governance reviewers. Proposed audience only; no Lookback recording, transcript or customer project is included.
TWO TIME HORIZONS
Trial today. A habit tomorrow?
Near term · 0–90 days
Over 0–90 days, give teams synthetic moderated and unmoderated sessions with salient clips, quieter counterexamples and generated suggestions. Compare source verification, observer convergence, overstatement and the quality of final evidence statements.
Longer term · 3–12 months
Over 3–12 months, follow approved review practices across projects and researchers. Study reuse of clips, context loss, duplicated findings, reviewer calibration and whether later evidence changes earlier interpretations.
What would make the result actionable?
Use scripted recordings and transcripts with known themes, counterexamples, timestamp links, observer-note histories and blinded adjudication. Any conclusion about real users requires consented research, a defensible sample and analysis appropriate to the study design.
A Gather simulation returns hypothetical customer reactions. Quantifying revenue, traffic or retention needs actual business inputs and validation against observed behavior.