On this page
- Structure is what predicts: structured interviews averaged .42 validity in Sackett and colleagues' 2022 meta-analysis, against .19 for unstructured ones.
- Mix question types: what would you do, what did you do, and show us.
- Write the rubric before you send the first link, and calibrate it with two reviewers.
- AI should transcribe and score against the rubric; a person decides.
What makes a video interview question predictive?
Three things: it tests a skill the job actually needs, every candidate gets the same question, and answers are scored against the same written standard. That is a structured interview, the most valid method in the largest recent meta-analysis of selection methods. Generic questions ("tell me about yourself") and gut-feel scoring are what make interviews unreliable. For how scoring works, see how AI video interviews work.
Which types of question should you use?
| Type | Format | Tests |
|---|---|---|
| Situational | "What would you do if…" | Judgement in the situations the job brings |
| Past behaviour | "Tell us about a time when…" | What the candidate has actually done |
| Job task | "Explain…", "Role-play…", "Walk us through…" | The skill itself, demonstrated |
| Knowledge | Short technical or process question | Must-have knowledge |
| Logistics | Notice, location, shifts | Basic fit (often better on WhatsApp) |
What are good AI video interview questions by role?
| Role | Question |
|---|---|
| Inside sales | "You have 60 seconds with a prospect who says they already use a competitor. Pitch us." |
| Inside sales | "Tell us about a target you missed. What did you change the next month?" |
| Customer support | "A customer was charged twice and is angry. Respond as you would on the call." |
| Customer support | "Explain how to reset a password to someone who has never done it." |
| Software developer | "Walk us through a bug you fixed recently: how you found it and how you knew it was fixed." |
| Software developer | "How would you design a URL shortener for 1,000 requests a second? Two minutes, high level." |
| BFSI operations | "A customer's documents don't match their application. What do you do?" |
| Team lead | "One of your best agents has started arriving late. How do you handle the first conversation?" |
How do you write the scoring rubric?
- List three to five must-have skills with the hiring manager.
- For each question, describe a strong, acceptable and weak answer in one or two lines.
- Add evidence to look for: specific actions, results, steps in the right order.
- Calibrate: two people score the same ten answers and agree where they differ.
- Review monthly: compare scores with who passed the next round and who performed well.
What settings work best?
- Four to six questions, 15 to 25 minutes in total.
- 30 seconds to think before recording, one to two minutes to answer.
- One re-record allowed, to reduce nerves and technical problems.
- Mobile-friendly, with a practice question first.
- A clear note on what is assessed and that a person reviews every answer.
Set up your questions and rubric
One-way video interviews scored against your rubric with quotes from each answer, and a person making every decision.
What are the most common mistakes?
- Generic questions that any candidate can rehearse.
- Too many questions, which lowers completion.
- Scoring presentation: faces, accents, background or "confidence". HireVue dropped facial analysis in 2021.
- No calibration, so scores depend on who reviews.
- Treating scores as decisions. Scores rank and explain; people decide.
Frequently asked questions
How many questions should an AI video interview have?
Four to six for a first round, taking 15 to 25 minutes in total.
Should candidates be allowed to re-record?
Allowing one re-record reduces stress and technical problems without changing what you learn.
Can AI score video interview answers fairly?
It can apply your rubric consistently and show the quotes behind each score. Fairness depends on job-related criteria, human review and never scoring faces, accents or tone.
The question types, examples and rubric steps were written for this guide using structured-interview research and how Hirin.ai's video interviews are set up, as of October 2026. Research was checked against its sources on 12 October 2026.
- Sackett, P. R., Zhang, C., Berry, C. M. and Lievens, F. (2022), Revisiting meta-analytic estimates of validity in personnel selection, Journal of Applied Psychology.
- SHRM, HireVue discontinues facial analysis screening (3 Feb 2021).
- Langer, M., König, C. J. and Krause, K. (2017), Examining digital interviews for personnel selection, International Journal of Selection and Assessment.
Rajni Bansal
Rajni Bansal is a seasoned HR leader with 15+ years of experience driving people strategy across global tech and services organizations. She brings deep expertise in talent management, digital HR transformation, and AI adoption in recruitment. As a contributor to Hirin.ai, Rajni shares practical insights on how HR teams can leverage emerging technology to build agile, future-ready workplaces.