How it worksPricingFAQBlogSign inStart free
Start free trial

How hiring managers should use an AI interview summary

How to read an AI interview summary, when to trust it, when to watch the full answer, and how to turn it into fair, fast hiring feedback.

DeepHire Team

8 min read

Key takeaways

  • Use an AI interview summary to decide which answers to watch first, not to make the hiring decision.
  • Read the rubric before the summaries, because a summary only measures what the rubric asks for.
  • Write your own rating before you look at the AI score so the score does not anchor you.
  • Watch the full answer before rejecting anyone on a role-critical question or near your cutoff.
  • Send feedback that points to specific answers and notes where you disagreed with the AI score.

An AI interview summary is a short written recap of what a candidate said in a recorded interview, usually broken out by question and often paired with a suggested score against a rubric. Hiring managers get the most from one when they treat it as a guide to which answers to watch, not as the decision itself.

That sounds like a small distinction, but it changes how you spend your review time. A summary can turn a stack of twenty recorded interviews into a sensible review order. It cannot tell you whether a candidate's judgment fits your team, and it can get details wrong.

What does an AI interview summary actually give you?

A typical tool transcribes the recorded answers and condenses each one into a few sentences. It then checks what was said against criteria the recruiter set up for the role. Some also give a score per criterion or an overall score.

That is useful. In thirty seconds you can see that a candidate walked through a specific customer escalation, including the steps they took and how it ended. You can also see when an answer stayed vague.

Summaries fall short in a few predictable ways:

  • They work from words, so they miss tone and how a candidate handles a hard question in the moment.
  • They can mishear names and technical terms, especially with background noise or a speaker the transcript struggles with.
  • They do not know your team or the unwritten parts of the job.
  • They reflect the rubric they were given. A weak rubric produces a confident-sounding summary of the wrong things.

For the bigger picture of where this fits in screening, see how AI candidate screening works.

What should a good summary include?

Before you rely on summaries, check that yours contain the parts below. If something is missing, ask your recruiter whether the rubric or the setup can change.

Part of the summaryWhat it helps you doWhat to check yourself
Per-question recapSkim answers without playing videoSpot check one answer against the recording.
Specific evidence quoted or paraphrasedSee the example the candidate usedMake sure the example is real and relevant, not stretched by the summary.
Score per rubric criterionFind strong and weak areas fastDecide whether you agree with the criterion itself.
Gaps or missing informationKnow what to ask in the next roundConfirm the candidate left it out and the transcript did not drop it.
Overall score or rankingSet a review orderNever use this alone to reject.

The most useful line in a good summary is usually the evidence, meaning the actual example the candidate gave. You can check that against the video. A score is only a shorthand for it.

How to read an AI interview summary: a five-step routine

The routine does not need to be long. It needs to be the same every time, so every candidate is compared the same way.

1. Read the rubric before the summaries

Look at what the summary is scoring against. Say the rubric asks the candidate to "explain a time they handled an upset customer," but your real concern is whether they can work alone on weekend shifts. In that case the summary will not help you, so fix the rubric with your recruiter first. A clear scorecard makes this much easier, and the hourly candidate scorecard guide has a template you can adapt for most roles, not only hourly ones.

2. Skim every summary once, without scores

Read the recaps and evidence for each candidate, and leave the scores for later. Note which candidates gave concrete examples and which stayed general.

3. Pick the answers worth watching

Choose one or two answers per candidate to watch in full. Start with the question most tied to success in the role. If you have room for a second, pick an answer where the summary seems unclear or one that will decide yes or no.

4. Watch those answers at faster speed

Answer length depends on the time limit your recruiter set. If answers are capped at three minutes, watching two answers each for eight candidates at 1.5x takes about half an hour, often less. Watch for what a summary cannot capture, such as whether they sound like they did the work and how they talk about other people.

5. Record your own rating, then compare

Write down your rating before you look at the AI score. Where you and the score agree, move on. Where you disagree by a lot, rewatch the answer and write one sentence about why. Those disagreements are the most useful thing you can send back to your recruiter, because they show where the rubric or the summary is off.

When should you skip the summary and watch the full interview?

Some situations call for the full video every time:

  • Any candidate you are about to reject on a role-critical criterion.
  • Scores near your cutoff, where one answer could swing the decision.
  • A summary that seems much stronger or weaker than the rest of the candidate's application.
  • Recordings with poor audio or speech the transcript may have struggled with.
  • Roles where spoken communication is part of the job, such as sales or front desk work.
  • Final shortlist candidates, before you commit live interview time to them.

On most shortlists this adds only a handful of full interviews. In exchange, you avoid rejecting a strong candidate because a transcript dropped half their answer.

Mistakes hiring managers make with AI summaries

  1. Treating the score as the decision, when it is only advice. A person should make the call and be able to explain it in plain terms tied to the job.
  2. Comparing scores across different roles. A 4 on a warehouse lead rubric and a 4 on an account manager rubric measure different things, so only compare candidates who were scored on the same rubric.
  3. Watching full video only for the candidates you already like. If you watch full answers for people you want to advance but rely on summaries to reject everyone else, you are applying two standards. Use the same review steps for everyone on the shortlist.
  4. Copying the summary into your feedback. Your recruiter already has the summary, so what they need from you is your own judgment.
  5. Asking the tool to judge things outside the rubric. Vague ideas like "culture fit" or "energy" invite inconsistent and unfair calls. If something matters, define it as a job-related behavior and add it to the rubric.

How to give useful feedback using summaries

Good feedback is short and points at evidence. Paste this format into your shared comments for each candidate:

  • Decision: advance, hold, or decline.
  • Strongest evidence: the one answer that most supports your decision, with the question number.
  • Concern: what you are unsure about, tied to a rubric criterion.
  • Where I disagreed with the AI score: criterion, my rating, one-sentence reason.
  • Ask in next round: one question to test the concern.

Compare two versions of the same feedback. The weak version reads: "Seems good, summary looks strong, move forward." The useful version reads: "Advance. Q2 answer on the double-booked delivery was specific and she owned the mistake. Concern on scheduling software, Q4 was vague. I rated Q4 a 2, AI gave a 4, and the summary missed that she has only watched others use it. Next round: ask her to walk through building a weekly schedule."

The second version takes about a minute longer to write. It also tells your recruiter that Q4 scoring needs a closer look for other candidates.

In DeepHire, AI summaries and scores are based on the recruiter's own rubric and are advice only. A hiring manager can open the shortlist link a recruiter shares without creating an account. Whatever tool your team uses, ask to see the summary and rubric next to the video so you can check one against the other.

Fairness, disclosure, and the rules

Rules on AI in hiring vary by location, and they change. Depending on where you hire, you may need to tell candidates when AI is used or have certain tools audited for bias. Other places limit how automated scores can feed into a decision. Check with your legal team before you change how your team uses AI scores.

A few habits help no matter where you hire:

  • Keep a person as the decision maker and write down the job-related reason for each decision.
  • Use the same rubric and the same review steps for every candidate in a role.
  • Tell candidates plainly that their answers may be summarized by AI. This guide on AI video interview disclosure has wording you can adapt.
  • Ask your recruiting team whether summaries and scores are checked for uneven results across groups. A bias audit for AI hiring tools explains what that check looks for.
  • Pass accommodation requests to your recruiter right away, so a candidate who needs another format is not scored on a recording they could not make in the usual way.

Say you notice one group of candidates getting lower scores on a criterion while their videos look comparable. Tell your recruiter. You are the one watching the recordings, so you may be the first person in a position to see it.

Frequently asked questions

Are AI interview summaries accurate?

Accuracy depends on the tool and the audio. Summaries can mishear names and technical terms, or drop part of an answer when the recording is noisy. Spot check one answer per candidate against the video, and always watch the full answer before rejecting someone on a key question.

Can a hiring manager reject a candidate based only on an AI score?

It is not good practice. Treat the score as advice and watch the relevant answers yourself, then record a job-related reason for the decision. Rules on automated decisions vary by location, so check with your legal team.

Do candidates need to be told their interview will be summarized by AI?

In some places it is required, and it is good practice everywhere. A short, plain note in the interview invite is usually enough. It should say that AI may summarize answers and that a person makes the decision. Confirm the requirements with your legal team.

How long should it take a hiring manager to review summarized video interviews?

Skimming the summaries for a shortlist takes a few minutes. Watching one or two key answers per candidate at faster speed takes longer. For a shortlist of around eight candidates, the whole review, including written feedback, often fits in under an hour.

Screen your next shortlist this week

Send one link. Candidates record on their own time. You review the whole batch in an afternoon.

Start your free trial

Free for 14 days. No credit card.