TL;DR: A structured interview question is a job-related question asked in the same wording and order to every candidate and scored against a written rubric. In AI and asynchronous interviews each question must stand on its own, because there is no live clarification, and each rubric anchor must be explicit. This page gives structured interview questions by role with 1-5 anchored rubrics.
Updated 21 October 2026 · Bank v1.0 This page is guidance, not legal advice.
Key takeaways
- A structured interview asks every candidate the same questions in the same order and scores them against anchored criteria. Greenhouse describes it as designed to assess candidates against the criteria for a specific job, with defined questions and a rubric; Workable describes asking each candidate the same questions in the same order; HR Cloud uses a scored scale.
- AI and async delivery change five things: questions must stand alone; answer length and think time need limits; the scorer is a model, so rubric anchors must be explicit; core questions stay fixed for comparability; candidates should be told the format and that AI is used.
- Use a 1-5 anchored scale (1 = no evidence, 3 = adequate, 5 = exceptional with specific examples) and a golden-answer key for every question.
- Structure is the active ingredient in the validity research. According to Sackett et al. (2022), structured interviews had an operational validity of .42 against .19 for unstructured ones, and SIOP's TIP reads the paper as saying structured interviews "may in fact be the strongest predictor of job performance" (SIOP TIP). Both figures are for human-rated interviews.
The structured-interview guides we read all assume someone is in the room. In an AI or asynchronous interview nobody is: the candidate cannot ask what a question means, and the scorer can only apply the rubric it was given. That makes loose wording and vague anchors expensive. This page is the version built for that setting. Use it to run the same question set through an AI interview platform or any async tool, with a 1-5 rubric and a golden-answer key attached to every question.
What is a structured interview question?
Definition. A structured interview question is a job-related question asked in the same wording and, as far as possible, the same order to every candidate, and scored against a written rubric. It tests one competency, comes with a defined set of follow-up probes, and is rated before any overall judgement is formed.
Greenhouse groups structured questions into four types: verification, behavioral, situational and skills test. Workable describes the common rating conventions: a five- or seven-point scale or pass/fail, with the STAR format for behavioral questions. HR Cloud uses a 1-5 scale, anchoring 1 as no evidence, 3 as adequate and 5 as exceptional with specific examples. (Greenhouse updated 27 April 2026; Workable modified 26 September 2023; HR Cloud published 3 March 2026. We fetched all three on 15 October 2026.)
Those three guides describe structured interviewing for a live interviewer. In the articles we read, none combines per-role question sets, anchored rubrics and AI or asynchronous delivery. This page does.
Why structure matters is a separate body of evidence. Why structure predicts performance sets it out in full, so this page does not repeat it.
What changes when the interviewer is an AI or the format is async?
Five things change. Plan for each before you write a single question.
- Write questions that stand alone. There is no live clarification, so name the situation and state what the answer should cover.
- Limit answer length and think time. Set both per question, so every candidate has the same room to answer.
- Write explicit rubric anchors. The scorer is a model. It needs "golden answer" key points, not an interviewer's impression. How the scorer uses the key you write is covered in how AI interviews are delivered and scored.
- Keep the core set fixed. Adaptive follow-ups can sit on top, but every candidate answers the same core questions, so scores compare.
- Tell candidates the format and that AI is used. Disclosure expectations differ by place; NYC Local Law 144 and the EU AI Act both contain notice or transparency provisions, so check the official text with counsel.
Fixed core questions, adaptive follow-ups
Ask every candidate the same core questions in the same order. A follow-up can then probe the missing part of an answer, such as the outcome or the candidate's own action, without opening a new topic. Define the probe set for each question in advance. Whether a given tool supports adaptive follow-ups varies, so treat this as guidance for any tool.
How do you write a question that stands on its own?
Work through this checklist for every question.
- Name the situation. Give the context the candidate would otherwise ask for.
- State what to cover. Tell them to describe the situation, their action and the result.
- Remove jargon only a live interviewer could explain. If you would have to unpack a term, replace it.
- Set the answer time. Say how long the candidate has.
- Test one competency. Split any question that measures two.
Our recommended answer times are below. They are our guidance, not sourced benchmarks, so adjust them to the role.
| Question type | Suggested answer time |
|---|---|
| Verification (facts about experience or licences) | 30-60 seconds |
| Situational ("what would you do if...") | 90 seconds to 2 minutes |
| Behavioral (a past example) | 2 minutes |
| Skills explanation ("walk through how you would...") | 2 to 3 minutes |
Three before-and-after rewrites show the change.
| Live-interview version | Standalone version | What changed |
|---|---|---|
| "Tell me about a difficult customer." | "Describe one occasion when a customer was unhappy about something you could not change. Say what the situation was, what you did, and what happened afterwards. You have two minutes." | "Difficult" is undefined. The rewrite sets the situation, the three parts to cover and the time. |
| "Walk me through your process." | "Describe how you prepare for a first call with a new prospect, from receiving the lead to dialling. Name the steps in order. You have two minutes." | "Your process" has no start or end. The rewrite fixes both and asks for an order. |
| "What would you do if throughput dropped?" | "A packing line that usually handles 120 orders an hour falls to 80 halfway through a shift. List the first three things you check, in order, and say why. You have 90 seconds." | "Throughput" and "dropped" depend on context. The rewrite gives numbers, a count and a time limit. |
How do you score answers? The 1-5 anchored rubric
Score every answer on a generic 1-5 scale. HR Cloud's 1 / 3 / 5 anchors are the only anchored scale we found among the guides we read, so we kept them and added 2 and 4.
| Score | Anchor |
|---|---|
| 1 | No evidence. No example, or a general claim with no action or result. |
| 2 | Some evidence, but weak. An example is named, but the action is unclear or the result is missing. |
| 3 | Adequate. A specific example and at least one concrete action, with a stated result. |
| 4 | Good. A specific example, several concrete actions and a clear result, with some reflection. |
| 5 | Exceptional, with specific examples. Every key point is covered in detail, and the candidate states what they would change. |
Use the generic 1-5 scale for every question in this bank. Do not mix it with another scoring scale.
One fully worked question
Competency: customer problem-solving under constraints. Answer time: 2 minutes.
Question: "Describe one occasion when a customer was unhappy about something you could not change. Say what the situation was, what you did, and what happened afterwards. You have two minutes."
Golden-answer key points
- States a specific situation and what was outside the candidate's control.
- Acknowledges the customer's problem before explaining the limit.
- Takes at least one concrete action, such as offering an alternative or escalating with context.
- Reports the outcome and what the candidate would do differently.
| Score | Example anchor for this question |
|---|---|
| 1 | "I always stay calm with customers." No situation, action or outcome. |
| 2 | Names a late delivery, says "I apologised." No further action or result. |
| 3 | Names a late delivery and the carrier delay, offers a replacement date, and says the customer accepted. |
| 4 | As 3, plus acknowledges the customer's missed event first, confirms the date in writing, and notes the outcome. |
| 5 | As 4, plus states what was said and when, who was looped in, and what the candidate now does earlier in the process. |
Strong-answer signals: a named situation, a sequence of actions in order, an outcome the customer or team could confirm, an honest limit. Red flags: blames the customer; describes a trait instead of an action; no outcome; an example where nothing was outside their control.
Bias-risk note: score the answer, not the delivery. Do not score accent, fluency, speaking pace, camera set-up, background or the type of customer named.
The question bank by role
These are sample questions. Each question is written to stand alone, tied to a competency, and carries a rubric and a bias-risk note. The questions are written by the ZeaHire team, not sourced from competitor lists, and the bank is versioned (v1.0). It contains no questions about protected characteristics. Tie each question to a named competency or skill, as skills-based hiring through interviews recommends, so you can show why it is job-related.
Table columns: competency · question · answer time · rubric anchors (1 / 3 / 5) · key points · follow-up probe · bias-risk note.
Customer support
| Competency | Question | Time | Anchors (1 / 3 / 5) | Key points | Follow-up probe | Bias-risk note |
|---|---|---|---|---|---|---|
| De-escalation | "A customer writes in angry that a refund has not arrived after ten days. Tell us what you would say in your first reply and why." | 90 sec | 1: generic apology, no next step. 3: acknowledges, gives one next step. 5: acknowledges, gives a next step with a timeframe, explains why. | Acknowledge; check status; set a timeframe; offer escalation. | "What if a policy blocks the refund?" | Do not score writing style or accent. |
| Policy accuracy | "You are not sure whether a customer qualifies for a warranty repair. Describe what you do before you answer them." | 90 sec | 1: guesses or promises. 3: checks policy or asks a colleague. 5: checks written policy, records the case, tells the customer what happens next. | Verify first; do not promise; document; set expectations. | "Where do you look first, and who do you ask second?" | Do not score product jargon. |
| Prioritisation | "You start a shift with five open tickets: two billing, one outage report, two how-to questions. Give the order you answer them and why." | 90 sec | 1: no order or reason. 3: an order with one reason. 5: an order that weighs impact and urgency, with a reason for each. | Outage first; billing by deadline; how-to last; reasons stated. | "What changes if a how-to customer is a key account?" | Do not score English fluency beyond clarity of content. |
Sales
| Competency | Question | Time | Anchors (1 / 3 / 5) | Key points | Follow-up probe | Bias-risk note |
|---|---|---|---|---|---|---|
| Discovery | "Describe how you prepare for a first call with a new prospect, from receiving the lead to dialling. Name the steps in order." | 2 min | 1: no steps. 3: several steps in order. 5: ordered steps that include research on the buyer's problem and a call goal. | Research; goal for the call; opening question; next step. | "What do you skip when you have ten minutes?" | Do not score vocal energy or charisma. |
| Objection handling | "A prospect says your price is too high. Tell us the first two things you would say or ask, and why." | 90 sec | 1: concedes or argues. 3: asks one clarifying question. 5: clarifies what "too high" means, then ties value to the stated need. | Clarify; do not discount first; tie to need. | "What if they will not say what they compare it with?" | Score content, not confidence of delivery. |
| Pipeline discipline | "Describe how you decide which deals to work this week and which to leave." | 2 min | 1: "whichever feels hot." 3: a stated rule. 5: a stated rule with stage, value and last-contact criteria, and a review habit. | Criteria; review rhythm; letting deals go. | "Give an example of a deal you dropped." | Do not score industry-specific vocabulary. |
Software engineer
| Competency | Question | Time | Anchors (1 / 3 / 5) | Key points | Follow-up probe | Bias-risk note |
|---|---|---|---|---|---|---|
| Debugging | "A service that passed its tests last week now returns errors for some users. List the first four things you check, in order, and why." | 2 min | 1: no method. 3: plausible checks, little order. 5: ordered checks from cheapest to costliest, each with a reason. | Recent changes; logs; reproduce; isolate affected users. | "Which check would you drop if time were short?" | Do not score the tools named, only the reasoning. |
| Collaboration | "Describe one occasion when you disagreed with a colleague's technical approach. Say what the disagreement was, how it was resolved, and what you learned." | 2 min | 1: no example. 3: example and resolution. 5: example, resolution, and a changed practice. | Specific disagreement; evidence used; outcome; lesson. | "What would you do differently?" | Do not score assertiveness style. |
| Design trade-offs | "You must store and search ten million short text records. Name one approach, one reason you chose it, and one weakness." | 2.5 min | 1: no approach. 3: approach and reason. 5: approach, reason and a specific weakness with a mitigation. | One approach; stated reason; honest weakness. | "What would make you change your choice?" | Do not score familiarity with a particular vendor. |
Recruiter and staffing roles
Staffing and recruiter questions cover sourcing, client management and candidate care. To run the same bank across several clients, keep the recruiter questions fixed and swap only the role-specific block for each client. AI for staffing agencies covers that workflow.
| Competency | Question | Time | Anchors (1 / 3 / 5) | Key points | Follow-up probe | Bias-risk note |
|---|---|---|---|---|---|---|
| Sourcing | "You have a role with a rare skill combination and no applicants after a week. Describe the first three sourcing steps you take, in order." | 2 min | 1: waits or reposts. 3: three plausible steps. 5: three ordered steps with a reason and a way to measure each. | Rework the brief; widen sources; test the criteria. | "What would you ask the client to relax?" | Do not score network size or employer prestige. |
| Client management | "A client says every candidate you sent is a poor fit. Say what you would ask them and what you would change." | 90 sec | 1: defends the candidates. 3: asks for feedback. 5: asks for specific examples, then changes the brief and the screen. | Ask for examples; recalibrate; agree a check-in. | "What if the client cannot say what is missing?" | Do not score tone of voice. |
| Candidate care | "A candidate you submitted is rejected by the client with no feedback. Describe what you tell the candidate and what you do next." | 90 sec | 1: no contact. 3: tells them honestly. 5: tells them promptly, seeks feedback, and keeps them in the pool where appropriate. | Honesty; speed; follow-up with the client. | "What do you say if the candidate asks why?" | Do not score warmth of delivery over content. |
Warehouse and operations
| Competency | Question | Time | Anchors (1 / 3 / 5) | Key points | Follow-up probe | Bias-risk note |
|---|---|---|---|---|---|---|
| Safety | "You see a pallet stacked unsafely in a walkway. Describe what you do in the next five minutes." | 90 sec | 1: ignores or walks past. 3: reports it. 5: makes the area safe, reports it, and follows the procedure. | Protect people first; report; follow procedure. | "What if your supervisor is off site?" | Do not score physical attributes. |
| Process accuracy | "The shelf count for a fast-moving item does not match the system. Describe what you do, in order." | 90 sec | 1: adjusts the number. 3: recounts and reports. 5: recounts, checks recent movements, reports with detail, does not alter records unilaterally. | Recount; check movements; report; no silent edits. | "How would you tell if this happens often?" | Do not score prior employer or equipment brand. |
| Reliability | "Describe one occasion when you covered an unplanned gap in a shift. Say what was needed, what you did, and what happened." | 2 min | 1: no example. 3: example and action. 5: example, action, result, and how it was handed over. | Specific gap; action; result; handover. | "What did you give up to cover it?" | Do not ask about family or availability reasons. |
Healthcare
| Competency | Question | Time | Anchors (1 / 3 / 5) | Key points | Follow-up probe | Bias-risk note |
|---|---|---|---|---|---|---|
| Patient safety | "You notice a colleague has skipped a required check on a patient. Describe what you do and who you tell." | 2 min | 1: ignores. 3: tells someone. 5: acts to protect the patient, tells the right person, and follows the reporting route. | Protect the patient; report; follow policy. | "What if the colleague is your supervisor?" | Do not ask about the candidate's own health. |
| Communication | "Explain to a worried family member, in plain language, why a procedure has been delayed. Assume they have no medical background." | 90 sec | 1: jargon or no explanation. 3: plain explanation. 5: plain explanation, acknowledges worry, and gives a next step. | Plain words; empathy; next step. | "What if they ask for details you cannot share?" | Do not score accent or speaking pace. |
| Documentation | "Describe how you make sure your notes at the end of a shift let the next person continue safely." | 90 sec | 1: no method. 3: a routine. 5: a routine with handover checks and a way to flag open items. | Clear notes; handover; flag open items. | "What do you do when time runs out?" | Do not score handwriting or typing speed. |
How do you use the bank at volume?
Follow four steps.
- Pick the role bank. Start from the role closest to yours.
- Tailor two or three questions to the job. Swap in the situation your role actually faces and keep the rubric structure.
- Score every answer against the rubric before you form any overall judgement.
- Let a human review the evidence and decide. The score informs the decision; it does not replace it.
When the shortlist is the bottleneck, narrow it first. Our tutorial on screening 200 CVs with AI covers that step.
The bank works in any async or AI interview tool: copy the question text and key points into whichever one you use. ZeaHire's Interview Templates, for example, hold questions under a name, type, round and language. They can be AI-generated by count, difficulty, industry and skills. Each answer is scored with the reasoning visible to the reviewer. Whichever tool you pick, check that it shows the reasoning and lets you hold the rubric constant across candidates. AI interview software that scores against a rubric is the category to evaluate.
Where do structured interviews go wrong?
Each of these breaks comparability between candidates. Guard against each one.
- Write the probe set before the interview. Off-script probing with no defined probe set means candidates no longer face the same questions.
- Score each question before forming an overall view. Scoring on impression first contaminates the per-question scores.
- Anchor on content, not delivery. Rubric anchors that describe delivery reward style over substance.
- Tell candidates the format. Say it is structured, and say AI is used, if it is.
- Tie every question to the job. A question with no link to a job requirement is a question you cannot defend.
How this bank was built and verified
- Authorship. Questions, rubrics and rewrites are written by the ZeaHire team, not sourced from competitor lists.
- Version. v1.0, 21 October 2026. The bank is reviewed annually.
- Sources last verified. The Greenhouse, Workable and HR Cloud pages cited above were fetched on 15 October 2026.
- Scope. This is guidance, not legal advice.
Frequently asked questions
How many questions should an AI interview have?
Start from competencies, not a target count: pick the four to six the role truly needs and write one question for each. Then add up the answer times to check the total. We found no sourced benchmark for question count, so treat this as our guidance.
Should AI interview questions be adaptive or fixed?
Fix the core set and let follow-ups adapt. Every candidate answers the same core questions in the same order, so scores compare. Defined probes sit on top to recover missing detail.
How long should an asynchronous interview be?
Add the answer times of your questions, plus reading time for each. We found no sourced benchmark for total length. Keep it to what your competencies need.
How do you score open-ended answers consistently?
Use an anchored 1-5 rubric with key points for each question. Score every question before you form an overall view. Score the content of the answer, not how it was delivered.
What should a rubric anchor look like?
A rubric anchor should be concrete, observable and content-based. Write an example for 1, 3 and 5, as in the worked question above. Avoid anchors that describe tone, fluency or confidence.
Can AI generate good structured questions?
Yes, as a draft. Review every generated question for job-relatedness and non-discrimination before you use it. A human should also check that each question stands alone.
Are follow-up questions allowed in structured interviews?
Yes, from a defined probe set. Improvised probing weakens the structure, because candidates no longer face the same questions.
Do I need to tell candidates the interview is AI-led?
Tell them. Disclosure expectations vary by jurisdiction; NYC Local Law 144 and the EU AI Act both contain notice or transparency provisions. Have counsel check your candidate-facing notice.
Next step
Run one real role through ZeaHire. See an evidence pack in a demo.
Comments
No comments yet — start the conversation.