UX research generates massive amounts of qualitative data — interviews, usability sessions, survey responses — that traditionally takes weeks to analyze. AI changes this fundamentally, compressing synthesis from weeks to hours.
Where AI Transforms UX Research
- Interview analysis: Transcription + theme extraction from 20 interviews in hours
- Survey synthesis: Pattern identification across hundreds of open-ended responses
- Persona drafting: Structure research into actionable persona documents
- Research report writing: Turn notes into polished deliverables
- Study design: Generate discussion guides and survey questions
1. Research Planning and Study Design
Discussion Guide Creation
Prompt: Create a semi-structured interview discussion guide for researching
how freelance graphic designers manage client projects.
Research objectives:
1. Understand pain points in the client communication process
2. Identify current tools and workflows
3. Discover what's missing from their current solution
Format:
- Warm-up questions (3-4 rapport builders)
- Core topic questions (6-8 open-ended questions)
- Probe examples for each core question
- Closing questions
Style: Open-ended, non-leading, exploratory. Avoid yes/no questions.
Interview duration: 45 minutes.Survey Design
Prompt: Design a 10-minute survey to measure user satisfaction with our
B2B project management software. We want to understand:
1. Overall satisfaction (NPS)
2. Which features deliver most value
3. Where users experience friction
4. What's missing that they wish existed
Include:
- NPS question (standard format)
- Feature satisfaction (use a matrix with our 8 key features)
- 3 open-ended questions about friction and wish list
- Demographic questions (role, company size, usage frequency)
Flag any questions that might introduce bias.Usability Testing Script
Prompt: Write a usability testing script for testing our mobile banking app's
new "send money" flow with 5 participants.
Include:
- Facilitator intro and consent language
- Think-aloud protocol instructions
- 3 tasks to complete (send $50 to a contact, split a bill with 3 friends,
review transfer history)
- Observation notes template
- Post-task questions (ease of use, confusion points)
- Post-session reflection questions
Remote testing format (Zoom). Participants are ages 25-45, smartphone users,
not necessarily tech-savvy.2. Interview Analysis
Transcript Analysis
After running interviews and getting transcripts (Otter.ai, Rev.com):
Prompt: Analyze these 5 interview transcripts and identify:
1. Top 3-5 themes across all interviews
2. Notable quotes that best illustrate each theme
3. Contradictions or surprising findings
4. Unmet needs explicitly or implicitly mentioned
5. Emotional language indicating strong positive or negative reactions
Format your output:
- Theme name
- Frequency (how many participants mentioned it)
- Supporting quotes with participant ID
- Implications for product design
[PASTE TRANSCRIPTS]Affinity Diagram from Quotes
Prompt: I have 87 quotes from 12 user interviews. Organize them into
affinity groups based on underlying themes.
For each group:
- Create a theme name (verb + noun, e.g., "Struggling to track deadlines")
- List the quotes that belong there
- Write a 1-sentence synthesis of what this group reveals
[PASTE QUOTES WITH PARTICIPANT IDs]Single Interview Analysis
Prompt: Analyze this user interview transcript. The participant is a
35-year-old marketing manager who uses our email marketing tool.
Extract:
1. Jobs to be done (what they're trying to accomplish)
2. Pain points and friction
3. Workarounds they've created
4. Moments of delight or satisfaction
5. Vocabulary they use (exact words for features/problems)
6. 3 direct quotes I should highlight in my research report
[PASTE TRANSCRIPT]3. Survey Response Analysis
Open-Ended Response Synthesis
Prompt: I have 340 responses to this survey question:
"What would make this product better for you?"
Analyze these responses and:
1. Create a frequency-sorted taxonomy of requested improvements
2. Identify the top 10 most requested features/changes
3. Group similar requests even when worded differently
4. Flag any unusual or particularly insightful responses
5. Identify any safety/accessibility concerns mentioned
[PASTE RESPONSES]NPS Comment Analysis
Prompt: I have 200 NPS survey comments (scores 0-10 with open text).
NPS = 34. Promoters: 45%, Passives: 24%, Detractors: 31%
Separately analyze:
1. What Promoters (9-10) love most (top themes)
2. What Detractors (0-6) hate most (top themes)
3. What would convert Passives to Promoters
4. Compare: what's different between what promoters love and detractors hate?
[PASTE COMMENTS WITH SCORES]4. Persona Creation
Prompt: Based on these research findings from 20 user interviews and
150 survey responses with our productivity app users, create 3 user personas.
Research summary: [PASTE YOUR NOTES/FINDINGS]
For each persona include:
- Name and photo description (we'll source the photo)
- Demographic profile
- Job title and company context
- Goals (what they're trying to achieve)
- Pain points with current solutions
- How they currently use our product
- Technology comfort level
- A quote that captures their perspective
- Key insight for our design team
Format: One persona per section, ready for our research presentation.5. Research Synthesis and Reporting
Executive Summary
Prompt: Write a 1-page executive summary of this UX research study.
Study: 15 user interviews + 200 survey responses
Topic: New checkout flow for our e-commerce platform
Audience: Product leadership (non-designers, business-oriented)
Key findings from my analysis:
[PASTE YOUR BULLET POINTS]
Format:
- Research objective (1 sentence)
- Methodology (2-3 sentences)
- Top 3 findings with supporting evidence
- Design recommendations (3-5 specific recommendations)
- Next steps
Avoid UX jargon. Connect findings to business impact where possible.How Might We Questions
Prompt: Transform these research insights into "How Might We" questions
for our design sprint. HMW questions should be opportunity-focused,
not solution-prescriptive.
Insights:
1. Users abandon the form when they see 12 fields on one page
2. Users don't trust the site because they can't find security badges
3. Users want to save their progress but can't find the option
4. Mobile users struggle with the date picker component
5. Users are confused about which plan is right for them
Generate 2-3 HMW questions per insight. Vary the scope — some broad, some specific.Journey Map Draft
Prompt: Draft a customer journey map narrative for this scenario:
User: Small business owner trying to set up payroll for the first time
Our product: Online payroll software
Journey stage: From first awareness to first successful payroll run
For each stage (Awareness, Consideration, Trial, Onboarding, First Success), write:
- What the user is doing
- What they're thinking
- How they're feeling (emotion)
- Touchpoints with our product
- Pain points
- Opportunities for us to improve
Format as a table. Based on this research data: [PASTE YOUR FINDINGS]6. Automated Research with AI Tools
AI-Powered Research Platforms
Maze: Unmoderated usability testing with AI analytics Dovetail: Repository + AI theme extraction from interview notes UserTesting: Video sessions with AI highlight creation Hotjar AI: Session recording pattern analysis Sprig: In-product surveys with AI analysis
These tools handle the most time-consuming parts — transcription, initial coding, theme surfacing.
7. Competitive Research
Prompt: I'm conducting competitive UX research on onboarding experiences
for [product category]. Based on these screenshots and descriptions of
5 competitor onboarding flows:
[DESCRIBE EACH COMPETITOR]
Analyze:
1. Common patterns across competitors (industry conventions)
2. Differentiated approaches (where they deliberately diverge)
3. Apparent best practices in this category
4. Gaps and opportunities none of them are addressing well
5. What we should learn from or avoid replicatingBest Practices for AI in UX Research
Always verify: AI analysis of transcripts can miss nuance, irony, or context. Spot-check AI-identified themes against primary transcripts.
Preserve participant voice: Quotes should always come from actual participants, not be paraphrased by AI.
Maintain research ethics: Don’t use AI to generate fake participant responses or fabricate findings. Synthesis is fine; fabrication is not.
Use AI for scale, not shortcuts: AI is most valuable when you’ve done real research and need help finding patterns across large datasets — not as a replacement for talking to users.
The combination of AI-assisted analysis and solid research methodology produces better insights faster than either approach alone.