{"site":{"name":"Koji","description":"AI-native customer research platform that helps teams conduct, analyze, and synthesize customer interviews at scale.","url":"https://www.koji.so","contentTypes":["blog","documentation"],"lastUpdated":"2026-05-21T02:09:01.198Z"},"content":[{"type":"blog","id":"86bdb70e-47b1-49b4-ad94-09db225677db","slug":"empathy-interviews-complete-guide-2026","title":"Empathy Interviews: The Complete Guide to Designing User-Centered Products in 2026","url":"https://www.koji.so/blog/empathy-interviews-complete-guide-2026","summary":"A complete 2026 guide to empathy interviews — the qualitative interview methodology developed at the Stanford d.school for design thinking. Covers the four core principles (seek stories, ask why five times, embrace silence, notice contradictions), 25 proven question templates, the seven most common interviewer mistakes, empathy mapping for synthesis, and how AI-moderated voice interviews (like Koji) scale the practice from 8 conversations a quarter to 80 — with adaptive probing, real-time transcription, and automatic thematic clustering. 78% of UX and product teams now use AI in research workflows in 2026.","content":"# Empathy Interviews: The Complete Guide to Designing User-Centered Products in 2026\n\n**Short answer:** An empathy interview is a one-on-one conversation designed to elicit the **stories, feelings, and unspoken context** behind how someone experiences a problem. Originally developed by the Stanford d.school as the first stage of design thinking, empathy interviews are now used by product teams, educators, civic designers, and founders to ground decisions in real human experience. In 2026, AI-moderated empathy interviews — like those run on **[Koji](https://www.koji.so)** — let teams scale the practice from a handful of curated conversations a quarter to 20+ conversations a week, with thematic analysis and quote traceability built in.\n\nEmpathy interviews are not just \"user interviews.\" They are a specific methodology with a specific goal: not to validate, not to test, but to **understand what it feels like to be the person you are designing for**. Done right, they surface the gap between what people *say* they want and what they actually need. Done wrong, they produce flattering quotes that mislead the team into building the wrong thing.\n\nThis is the complete 2026 guide — methodology, questions, common mistakes, and the modern AI-moderated workflow.\n\n## What is an empathy interview?\n\nAn empathy interview is a structured, qualitative conversation rooted in the design thinking methodology developed at the Stanford d.school. Its goal is to **deeply understand a person''s experience** of a problem space — their stories, feelings, motivations, and contradictions — before any product, service, or solution is designed.\n\nThe d.school describes empathy as built on three practices: **immerse** (experience what your user experiences), **observe** (view users in the context of their lives), and **engage** (interact through scheduled interviews and short ''intercept'' encounters). The empathy interview is the *engage* part — the structured conversation that lets you ask, listen, and reflect.\n\nEmpathy interviews are distinct from:\n\n- **Usability testing** — which asks \"can you use this?\" (evaluative)\n- **Customer discovery interviews** — which ask \"is this problem real for you?\" (validation-driven)\n- **Jobs-to-be-Done interviews** — which ask \"what job did you hire this product to do?\" (motivation-driven)\n- **Concept tests** — which ask \"what do you think of this solution?\" (reaction-driven)\n\nEmpathy interviews ask: **\"What is it like to be you?\"**\n\nSee [Koji''s customer discovery interviews guide](/docs/customer-discovery-interviews) for the broader interview taxonomy.\n\n## Why empathy interviews matter in 2026\n\nThree structural shifts make empathy interviews more — not less — important in 2026:\n\n### 1. AI is generating output. Empathy generates direction.\n\nWith AI now capable of writing copy, generating UIs, and producing analysis on demand, the strategic question has shifted from \"how do we ship faster\" to \"are we shipping the right thing?\" Empathy interviews provide the **directional grounding** that AI cannot generate — the lived context behind a user''s problem. The teams who skip empathy in 2026 ship faster in the wrong direction.\n\n### 2. Research has democratized\n\nAccording to [The State of User Research 2025](https://www.userinterviews.com/state-of-user-research-report), 78% of UX and product teams now use AI in their research workflows — more than double the 34% adoption rate in 2024. The share of organizations where research is essential to all levels of business strategy has nearly tripled in one year, from 8% in 2025 to 22% in 2026. PMs, designers, and even marketers run their own studies now. Empathy interviews are the most accessible methodology for non-researchers to do well.\n\n### 3. AI moderation removes the moderator bottleneck\n\nThe historical objection to empathy interviews was scale — they require a trained moderator, hours of audio review, and weeks of synthesis. In 2026, AI-moderated voice platforms run empathy-style interviews 24/7 with adaptive probing, real-time transcription, and automatic thematic analysis. See [AI moderated interviews](/docs/ai-moderated-interviews) for the technical depth.\n\n## The Stanford d.school empathy interview methodology\n\nThe d.school methodology has four core principles, each with practical implications.\n\n### Principle 1: Seek stories, not opinions\n\nThe single most important rule: **ask for stories, not for opinions**. Opinions are post-hoc rationalizations. Stories carry the actual context, decisions, and emotions of a real moment.\n\n**Don''t ask:** \"What do you think of meal kits?\"\n**Ask:** \"Tell me about the last time you cooked dinner for someone.\"\n\nThe first question produces a survey-grade opinion. The second produces a story that reveals everything — who they cook for, what they buy, what frustrates them, what they are proud of, what they avoid.\n\nSee [how to write user interview questions](/blog/how-to-write-user-interview-questions) for question design depth.\n\n### Principle 2: Ask \"why\" five times\n\nThe d.school borrows \"Five Whys\" from Toyota''s manufacturing methodology. The first answer is rarely the real answer. Each \"why\" peels back a layer until you reach the underlying motivation, fear, or value.\n\n**Example chain:**\n- \"Why did you cancel your gym membership?\" → \"Too expensive.\"\n- \"Why was that the breaking point this month?\" → \"I had not been going.\"\n- \"Why had you stopped going?\" → \"It was hard to fit in after work.\"\n- \"Why is that hard?\" → \"I am too tired by 6pm.\"\n- \"Why too tired?\" → \"I have been sleeping badly since the baby came.\"\n\nThe cancellation was not about price. It was about life stage. The actual product opportunity (and the actual churn driver) is the fifth answer, not the first.\n\n### Principle 3: Embrace silence\n\nMost interviewers fill silence. Empathy interviewers sit with it. The d.school explicitly trains practitioners to **wait 3–5 seconds after an answer** before asking the next question. The participant often continues — and the continuation is where the most honest content lives.\n\n### Principle 4: Notice contradictions\n\nPeople are not consistent. They say one thing, then contradict it. They claim to value X while behaving as if they value Y. The empathy interviewer''s job is not to \"correct\" the contradiction — it is to **notice it, hold it, and explore it gently**.\n\nContradictions are not interviewer failures. They are research gold. They point to the gap between what people believe about themselves and what they actually do, which is where most product opportunity lives.\n\nSee [avoiding bias in interviews](/docs/avoiding-bias-in-interviews) for related interviewer skills.\n\n## How to plan an empathy interview\n\nEmpathy interviews look casual but require careful design. Plan in five steps.\n\n### 1. Define the experience you want to understand\n\nNot the product. Not the feature. The **experience**. Examples:\n- \"What is it like to manage finances as a freelancer?\"\n- \"What is it like to onboard a new hire as a remote-first team lead?\"\n- \"What is it like to learn a language as an adult immigrant?\"\n\nNotice that none of these mention a product. You are studying a human experience, not a product reaction.\n\n### 2. Recruit for diverse perspective, not statistical sampling\n\nEmpathy interviews are not surveys. You do not need 400 respondents. You need 5–8 people with **different relationships to the experience** — people in the middle of it, people who solved it, people who gave up on it, edge cases. See [how to recruit user research participants](/blog/how-to-recruit-user-research-participants-2026).\n\n### 3. Draft a question set, then throw away the script\n\nThe d.school methodology recommends drafting 8–12 questions, then treating them as **starting prompts**, not a script. The interview should follow the participant''s stories — sometimes covering questions 1, 3, 7 and skipping the rest, sometimes inventing new questions in the moment.\n\n### 4. Pair an interviewer with a notetaker (or use AI moderation)\n\nThe traditional empathy interview pair is one interviewer (asking, listening, probing) and one notetaker (capturing quotes, body language, observations). In 2026, an AI moderator like Koji replaces both roles — running the conversation with adaptive probing while transcribing in real-time and annotating themes. See [AI note taker for user interviews](/docs/ai-note-taker-user-interviews).\n\n### 5. Plan synthesis before you interview\n\nThe most common failure mode of empathy interviews is having 8 great conversations and no idea what to do with them. Plan synthesis up front. The d.school uses an **empathy map** (Says / Thinks / Does / Feels) or a **journey map**. Koji performs thematic clustering automatically across all interviews in a study — see [AI generated insights](/docs/ai-generated-insights).\n\n## 25 empathy interview questions that actually work\n\nThe d.school and adjacent practitioners have refined a question library over two decades. The questions below all share one property: **they invite a story instead of a judgment**.\n\n### Opening questions (build rapport)\n\n1. \"Tell me about yourself.\"\n2. \"Walk me through a typical day.\"\n3. \"What brought you to [the experience] in the first place?\"\n\n### Story-prompt questions (the core of the interview)\n\n4. \"Tell me about the last time you [did the thing].\"\n5. \"Tell me about the first time you [did the thing].\"\n6. \"Tell me about a time when [the thing] went really well.\"\n7. \"Tell me about a time when [the thing] went really badly.\"\n8. \"What is the longest you have ever spent doing [the thing]?\"\n9. \"Walk me through what happened the day before / day after [the moment].\"\n\n### Feeling questions (surface emotion)\n\n10. \"How did that feel?\"\n11. \"When you say that was frustrating — what does frustrating mean here?\"\n12. \"What were you afraid would happen?\"\n13. \"What did you hope would happen?\"\n14. \"Who else was affected by this?\"\n\n### Workaround questions (find unmet needs)\n\n15. \"What do you do when [the obvious solution] does not work?\"\n16. \"Have you ever tried something that surprised you that it worked?\"\n17. \"Have you ever recommended a solution to a friend who had the same problem? What did you tell them?\"\n18. \"What is the hack you use that you would never recommend out loud?\"\n\n### Belief questions (uncover values)\n\n19. \"What do you wish people understood about [the experience]?\"\n20. \"What do other people get wrong about this?\"\n21. \"What is the part you almost never tell anyone?\"\n\n### Contradiction questions (probe carefully)\n\n22. \"Earlier you said X, but just now you said Y — can you help me understand?\"\n23. \"If you could wave a wand and change one thing — what?\"\n\n### Closing questions (open the door for more)\n\n24. \"Is there anything I should have asked you that I did not?\"\n25. \"Who else should I be talking to?\"\n\nFor ready-to-use templates, see [customer interview questions templates](/blog/customer-interview-questions-templates) and [customer interview questions examples](/docs/customer-interview-questions-examples).\n\n## The 7 most common empathy interview mistakes\n\n### Mistake 1: Asking opinion questions instead of story questions\n\n\"Do you like X?\" produces opinions you cannot trust. \"Tell me about the last time you used X\" produces a story full of real data.\n\n### Mistake 2: Leading the witness\n\n\"Was that frustrating?\" plants the emotion. Ask \"How did that feel?\" and let the participant name their own emotion.\n\n### Mistake 3: Pitching your product mid-interview\n\nThe fastest way to destroy an empathy interview is to start describing your solution. The moment you do, the participant pivots from telling stories to giving you feedback on your idea. You will get politeness, not truth. See [the Mom Test](/blog/mom-test-customer-interviews-2026) for why this matters.\n\n### Mistake 4: Stopping at the first answer\n\nThe first answer is the rehearsed answer. The third answer is closer to the truth. The fifth answer is the truth. Ask why, then ask why again, then sit in silence.\n\n### Mistake 5: Solving instead of listening\n\nWhen a participant describes a problem, the interviewer''s instinct is to help — to suggest a workaround, to validate, to share their own story. **Resist all of it.** The job is to listen, not to participate.\n\n### Mistake 6: Skipping the contradictions\n\nPeople say one thing then do another. The temptation is to \"smooth over\" the contradiction. Don''t. The contradiction is the most valuable signal in the conversation.\n\n### Mistake 7: No synthesis plan\n\nEight great interviews, three weeks of audio, no insights. The most common failure mode in empathy research. Plan synthesis before you collect — or use a platform like Koji that synthesizes automatically.\n\n## Empathy mapping: turning interviews into insights\n\nThe d.school''s standard synthesis tool is the **empathy map** — a 2x2 (or sometimes 4-quadrant) framework that organizes interview content into four categories:\n\n| Says | Thinks |\n|---|---|\n| Direct quotes from the participant | Inferred thoughts and beliefs |\n| **Does** | **Feels** |\n| Observed behaviors and actions | Emotional state, tone, body language |\n\nAfter 5–8 empathy interviews, the team fills in the empathy map collaboratively, then identifies **patterns across participants** — moments where multiple people Says/Thinks/Does/Feels the same thing. Those patterns become the foundation for design opportunities.\n\nIn a 2026 AI-moderated workflow, Koji automatically clusters statements into Says/Thinks/Feels/Does buckets across every interview in a study, then surfaces the cross-participant patterns. The empathy map becomes the **output of the analysis**, not a manual workshop deliverable. See [AI transcript analysis guide](/docs/ai-transcript-analysis-guide).\n\n## How AI is changing empathy interviews in 2026\n\nFor 20 years the bottleneck on empathy interviews was scale: you could afford to run 8 conversations a quarter, not 80. That meant teams used empathy research for major strategic decisions and reverted to surveys for everything else. The shape of the practice was distorted by the cost of doing it.\n\nAI moderation changes that math. In 2026:\n\n- **Participants self-serve** into a study via a shareable link, complete the conversation on their own time\n- **An AI voice moderator** runs the interview with adaptive follow-ups, the Five Whys probing, and structured questions (Koji supports six: open_ended, scale, single_choice, multiple_choice, ranking, yes_no)\n- **Transcription happens in real time** with speaker diarization\n- **Themes cluster automatically** across the full study — Says / Thinks / Feels / Does, plus persona, segment, and behavior dimensions\n- **Reports generate in one click** with quote-level traceability\n\nThis does not replace the *strategic* practice of empathy research. It replaces the *operational* bottleneck — moderation, transcription, and synthesis — that historically capped the practice. A team that used to run 8 empathy interviews a quarter can now run 80, with the same one researcher overseeing methodology and synthesis.\n\nSee [continuous discovery user research](/docs/continuous-discovery-user-research) for how empathy interviews evolve into a continuous research habit.\n\n## When to use empathy interviews (and when not to)\n\nEmpathy interviews are the right tool when you need to:\n\n- **Understand a new problem space** before designing a product\n- **Diagnose unexpected user behavior** (low activation, surprising churn, unexplained drop-off)\n- **Build a shared mental model** of the user inside a team\n- **Ground a strategic decision** in lived experience, not stakeholder opinion\n- **Onboard new product, design, or research hires** to who the user actually is\n\nEmpathy interviews are the wrong tool when you need to:\n\n- **Validate a specific concept** (use concept testing instead)\n- **Measure something** (use a survey)\n- **Test a specific interaction** (use usability testing)\n- **Optimize a feature** (use A/B testing + analytics)\n\nThe empathy interview generates direction. Other methodologies validate, measure, or optimize within that direction.\n\n## Try AI-moderated empathy interviews\n\nKoji lets product, design, and research teams run empathy interviews at the speed of product cycles — not at the speed of consultant timelines. AI-moderated voice conversations with adaptive Five-Whys probing, six structured question types inside the same conversation, automatic empathy mapping and thematic clustering, and one-click stakeholder reports with quote traceability.\n\n**[Start free at koji.so](https://www.koji.so)** — and turn empathy from a one-off design sprint into a continuous research habit.","category":"Tutorial","lastModified":"2026-05-20T03:19:31.459627+00:00","metaTitle":"Empathy Interviews: The Complete 2026 Guide (Stanford d.school + AI Moderation) | Koji","metaDescription":"The complete 2026 guide to empathy interviews — Stanford d.school methodology, 25 proven questions, common mistakes, empathy mapping, and how AI-moderated voice interviews scale the practice for product teams.","keywords":["empathy interviews","empathy interview questions","design thinking interviews","user-centered design","stanford d.school empathy interview","empathy mapping","empathy interview methodology","ai empathy interviews"],"aiSummary":"A complete 2026 guide to empathy interviews — the qualitative interview methodology developed at the Stanford d.school for design thinking. Covers the four core principles (seek stories, ask why five times, embrace silence, notice contradictions), 25 proven question templates, the seven most common interviewer mistakes, empathy mapping for synthesis, and how AI-moderated voice interviews (like Koji) scale the practice from 8 conversations a quarter to 80 — with adaptive probing, real-time transcription, and automatic thematic clustering. 78% of UX and product teams now use AI in research workflows in 2026.","aiKeywords":["empathy interviews","design thinking","user research","customer interviews","d.school","empathy mapping","jobs to be done","ai interviews"],"aiContentType":"guide","faqItems":[{"answer":"An empathy interview is a structured one-on-one conversation designed to elicit the stories, feelings, and unspoken context behind how someone experiences a problem. Developed by the Stanford d.school as the first stage of design thinking, the goal is to deeply understand a user — not to validate a product idea or test a solution.","question":"What is an empathy interview?"},{"answer":"Empathy interviews ask \"what is it like to be you?\" — focused on understanding the human experience of a problem. Customer discovery interviews ask \"is this problem real and worth solving?\" — focused on validating a product opportunity. Empathy is generative; discovery is validation-driven.","question":"How is an empathy interview different from a customer discovery interview?"},{"answer":"The Stanford d.school methodology recommends 5–8 empathy interviews per study, with participants selected for diverse perspectives on the experience, not statistical sampling. With AI-moderated platforms like Koji, teams now run 20–80 empathy interviews per study without proportional cost or time increase.","question":"How many empathy interviews do you need?"},{"answer":"The strongest empathy questions invite stories, not opinions. Examples: \"Tell me about the last time you did X,\" \"Walk me through what happened,\" \"How did that feel?,\" \"What do you wish people understood about this?\" Avoid \"Do you like X?\" or \"Would you use Y?\" — those produce opinions you cannot trust.","question":"What are the best empathy interview questions?"},{"answer":"An empathy map is a synthesis framework with four quadrants — Says, Thinks, Does, Feels — used to organize content from empathy interviews. Teams populate the map with direct quotes (Says), inferred beliefs (Thinks), observed behaviors (Does), and emotional state (Feels), then look for patterns across participants. AI platforms like Koji can populate empathy maps automatically across an entire study.","question":"What is an empathy map?"},{"answer":"Yes — AI voice moderators in 2026 run empathy-style interviews with adaptive probing (including the Five-Whys methodology), real-time transcription, and automatic thematic clustering. The strategic skill of designing questions and interpreting findings still requires a human researcher, but the operational bottleneck — moderation, transcription, and synthesis — is now AI-handled.","question":"Can AI run empathy interviews?"}],"relatedTopics":["empathy interviews","design thinking","user research","customer interviews","d.school methodology","empathy mapping"]}],"pagination":{"total":1,"returned":1,"offset":0}}