Run Expert Interviews That Yield Article Material

To run an expert interview that produces usable article material, sequence the call from the reader's task to the expert's most recent concrete experience, and refuse to move on from any answer until it contains a specific: a real project, a number they actually saw, a step that failed, or a decision they made. The transcript of a good interview for an article reads like evidence. The transcript of a bad one reads like a keynote.

The difference is rarely the expert. It is the interviewer's sequencing and follow-ups. Experts default to safe generalities because generalities are always true and never embarrassing. Your job on the call is to make specifics feel safe and easy, then capture them precisely enough to survive editing, permission checks, and fact-checking.

Use the 60-minute run sheet below. It pairs with the expert interview content brief: the brief decides what the article must prove before you schedule the call; the run sheet is how you actually conduct the hour.

Who this is for

This is for writers, editors, content leads, and founders who interview practitioners — customers, engineers, operators, in-house experts — to produce articles with first-hand evidence. It assumes you have already decided the article deserves an interview and you have roughly an hour of the expert's time.

It is not media training, and it is not a promise that interview-led articles rank better. It is a working method for getting a transcript worth drafting from. Google's people-first guidance is the reason the method matters: the finished page has to demonstrate real experience a reader can use, and that experience has to exist in your notes before it can exist on the page.

How should you structure a 60-minute expert interview?

Structure the hour in six segments: permissions and recording consent, reader-task framing, one recent-work anchor story, depth passes on the story's decision points, contrarian and boundary questions, and a wrap that flags the quotes you expect to use. Copy the run sheet below into your notes doc and run the call against it. Times are a default, not a law; the sequence matters more than the clock.

Segment	Time	What you are doing	Exit condition
Setup and permissions	0–5 min	Confirm recording, how they want to be attributed, and what is off limits	Expert has agreed to recording and a default attribution level
Reader task framing	5–10 min	Describe the reader and the task the article serves; ask what people get wrong	Expert is talking about the reader's problem, not their own bio
Recent-work anchor	10–25 min	"Walk me through the last time you did this" — one real project, start to finish	You have one concrete story with steps, decisions, and at least one failure
Depth passes	25–45 min	Revisit the story's decision points; push for tradeoffs, numbers, and rejected options	Each planned article section has at least one specific attached
Contrarian and boundary questions	45–55 min	Where does the standard advice break? What would they never do?	You have at least one opinion the expert owns, with reasoning
Wrap and clearance	55–60 min	Flag quotes you expect to use, confirm follow-up channel, ask what you should have asked	Expert knows what happens next and what they will get to review

The exit conditions are the artifact's point. If a segment has not hit its exit condition, stay in it and drop a later question. An interview that produces one fully-specified story beats one that skims ten topics.

What order should you ask expert interview questions in?

Ask in this order: the reader's task first, then the most recent real example, then the decision points inside that example, then tradeoffs, and boundaries last. Question order does more work than question wording. (For the questions themselves, organized by what each article section needs, use the subject-matter-expert question bank; this section is about sequencing them.)

1. Start with the reader's task, not the expert's opinion. "When someone searches this, what are they actually trying to do?" This calibrates the expert to the audience and stops the interview from drifting into industry commentary.
2. Anchor on the most recent real instance. "Tell me about the last time you did this" outperforms "how do you usually do this." Recent and singular beats typical and averaged, because averaged answers delete the details you need. Nielsen Norman Group's interviewing guidance makes the same point about user interviews: people report specifics more reliably than they report habits.
3. Interrogate the decision points inside the story. Once you have a story, walk back through it: why that option, what else was on the table, what did it cost, when did you know it was working or failing.
4. Ask for boundaries last. "When would this advice be wrong?" and "what would you never do here?" produce the caveats and tradeoffs section of the article. Asked too early, they get hedges; asked after a story, they get rules with reasons.

Avoid leading questions at every stage. "Was that because of X?" hands the expert your hypothesis and most will politely accept it. Ask "what caused that?" and let them fill the gap. NN/g's discussion-guide advice is blunt about this: the guide exists to keep you from steering the participant toward what you hope to hear.

What follow-up questions turn generalities into specifics?

Use follow-ups that demand an instance: the most recent example, the first concrete step, the rough number, the rejected alternative, the time it failed. Follow-ups are where usable material actually appears. Keep these eight prompts printed next to your notes and use one every time an answer is abstract:

• "Can you give me the most recent example of that?"
• "What did that look like concretely — what did you actually do first?"
• "Roughly what numbers are we talking about?" (then note the roughness — a hedge is part of the fact)
• "What did you try before that, and why did you drop it?"
• "Who disagreed, and what was their argument?"
• "How did you know it was working?"
• "What would a smart person still get wrong after hearing that?"
• "You said it 'usually' works — tell me about a time it didn't."

The specificity check is simple: after each answer, ask yourself whether a fact-checker could challenge it. "Onboarding matters" cannot be challenged, so it cannot carry an article. "We cut the onboarding form from eleven fields to four and support tickets about setup dropped within a month" can be challenged — which is exactly what makes it usable, once you verify you may publish it and how precisely they measured it.

Silence is also a follow-up. Reuters' interview guidance treats preparation and accurate capture as the interviewer's core duties. The cheapest capture technique we know is silence: wait three seconds after the expert stops talking — the second half of an answer, produced to fill your silence, is usually more specific than the first.

How do you capture quotes and evidence during the call?

Capture during the call, not after: record with consent, timestamp highlights live, repeat numbers back, ask for artifacts in the moment, and tag claims that need verification. A great answer you cannot quote precisely is a paraphrase at best. In-call capture habits:

• Record, with consent, every time. Ask at minute zero, and confirm the attribution default (on record, background, or approval-before-publish). You will settle the details later in the quote permission workflow, but the default has to exist before anyone says anything interesting.
• Timestamp the highlights live. When an answer is quotable, type the timestamp and three words. Skimming a transcript for six markers is a ten-minute job; rereading the whole thing is an afternoon.
• Repeat numbers back. "Eleven fields down to four — did I get that right?" catches mishearing while the expert is still there, and signals that you intend to be accurate.
• Ask for artifacts in the moment. When the expert mentions a checklist, dashboard, or doc: "could you share a screenshot or a redacted version?" Asking during the call gets a yes far more often than an email three days later.
• Mark claims that need verification. Anything about a third party's product, policy, pricing, or performance gets a "verify" tag in your notes. The expert's memory is a lead, not a source — the sourcing workflow decides what the article may assert.

Mistakes that quietly ruin interview material

• Sending the full question list in advance. Send the topic and the reader task, not the questions. Prepared answers converge on the expert's existing talking points; specifics come from retrieval, not rehearsal.
• Accepting the composite anecdote. "Typically a client will..." is a fictional average. Redirect: "give me one real client, anonymized however you need."
• Filling silence with your own expertise. Interviewers who demonstrate knowledge get agreement instead of information. Spend your credibility on better follow-ups.
• Chasing the whole question list. Coverage is the enemy of depth. The run sheet's exit conditions exist so you drop questions, not specifics.
• Treating the expert as a citation machine. Interviews yield experience and judgment. Claims about tools, policies, or prices still need primary sources; do not let a confident quote substitute for one.
• Skipping the permission conversation because it feels awkward. Thirty seconds at minute zero and one minute at wrap costs nothing. Discovering at publish time that your best quote is off the record costs the article.

Claim ledger

Claim	Source	Status	Review rule
Interview guides should avoid leading questions and keep the conversation focused on what the interviewer needs to learn.	Nielsen Norman Group interviewing-users guidance	verified on 2026-09-13	Recheck within 180 days
People report specific recent instances more reliably than generalized habits, so recent-example questions produce better material.	Nielsen Norman Group user-interview guidance	verified on 2026-09-13	Recheck within 180 days
Interviews require preparation and accurate handling of quotations and attribution.	Reuters Handbook interview guidance	verified on 2026-09-13	Recheck within 180 days
Helpful, people-first content should demonstrate first-hand experience where relevant.	Google Search Central helpful content guidance	verified on 2026-09-13	Recheck within 90 days or if guidance changes
The run sheet and prompt bank are our own editorial artifacts, not measured performance claims.	Our own, dated 2026-09-13	disclosed	Present as workflow advice only

FAQ

How long should an expert interview for an article be?

Sixty minutes is the practical default: five for setup and permissions, twenty for one anchor story, twenty for depth passes, ten for boundaries, five for wrap. Thirty minutes works only when the article needs one section's worth of material and you skip breadth entirely.

Should I send interview questions in advance?

Send the reader task and the topic areas, not the verbatim questions. Advance questions produce rehearsed talking points; the usable material comes from live retrieval of real events plus your follow-ups.

What if the expert only gives generic answers?

Anchor harder on recency: "the last time you did this, what happened first?" If three attempts at a real example produce nothing concrete, the problem is the source, not the questions — capture their judgment as clearly-labeled opinion or find a second expert, rather than dressing generalities up as experience.

Can I run the interview by email instead of a call?

Email produces polished position statements and no follow-ups, which is the opposite of what articles need. Use async only for verification and quote approval after a live conversation, not for the conversation itself.

How is this different from the interview content brief?

The brief is the pre-call planning sheet: what the article must prove, which claims are source-sensitive, which quotes will need permissions. The run sheet is the in-call method. Fill the brief first, then run the hour against the run sheet, then hand the output to the quote clearance workflow.

Where this fits in the interview pipeline

Recruit the right person for the call with how to find experts worth interviewing when you don't have famous ones, and prepare it with the expert interview content brief. Afterwards, route everything you captured through the quote permissions, attribution, and fact-checking workflow before any of it is published, and draft from the transcript with the voice-preservation drafting protocol so the specifics you captured survive AI-assisted assembly. If the conversation was rich enough for several pieces, use the interview-to-article splitting map. For the standards underneath all of this, read what experience means in a blog post when you are not a famous expert.

Sources and last-reviewed notes

Last reviewed: 2026-09-14. Sources checked for this guide: Nielsen Norman Group guidance on interviewing users and running user interviews, the Reuters Handbook guidance on interviews, and Google Search Central guidance on helpful, people-first content. The 60-minute run sheet, exit conditions, and follow-up prompt bank are our own editorial artifacts; the example answers in this guide are illustrative phrasing, not real quotes, and no interview outcomes are claimed as measured results.