InsightHireHelp Center

AI interview fairness

How candidates are told about AI, how they can ask for a human review, and the safeguards around AI interview scores.

InsightHire uses AI to transcribe and score video interview answers. This page explains what candidates see, what they can choose, and the safeguards your team can rely on. Settings live in Settings → Responsible AI and need an Org Admin.

What candidates are told

The first time a candidate opens a journey or assessment, they see a privacy consent screen (titled Before You Begin unless InsightHire has published newer wording). It explains data collection and storage, video and audio recording, AI-assisted analysis, and third-party processing, and lists the candidate's rights. They tick the required boxes and choose I Agree — Continue.

If they choose Decline, InsightHire explains that they can't continue and no data will be collected, and asks them to Confirm Decline or Go Back. Each consent is recorded against the published wording they saw.

Candidates never log in for any of this. The journey link itself grants access.

AI notice

Before the first step, the welcome screen shows a short, plain-language notice that AI is used to help review answers. The notice is always shown, whatever you pick below. Each time it is shown, InsightHire records it in your disclosure activity.

"Prefer a human review"

Under the notice, candidates can tick Prefer a human review. It is unticked by default, so starting stays one click. If a candidate ticks it:

  • Their answers are still transcribed so your team can read them.
  • AI scoring and video analysis are skipped. Answers are marked Needs review.
  • A score gate never fails them automatically; your team decides.
  • Their application is flagged as having requested human-only review. Resume scoring is skipped, and AI avatar steps are blocked until a recruiter arranges an alternative. They are never called for an AI phone screen; on an AI Phone Screen journey step they record video answers instead.

Candidates can also ask for a human review, or withdraw consent for AI, at any time from /my-data. See Privacy requests and retention.

Choose when to offer it

In Settings → Responsible AI → "Prefer a human review" for candidates, pick one:

  • Automatic: only where the law requires it (default). For each invite, InsightHire checks the job's location, and the candidate's location if we have it. The option is offered for jobs in the EU and EEA, the UK, Colorado, New York City and Illinois. If the location is missing, remote or not recognised, it is offered, to be safe.
  • Always offer. Every candidate can choose a human review.
  • Never offer. The AI notice is still shown, but the option is hidden. This may break local rules for jobs in the places listed above, so check with counsel first.

The decision is made on our servers, not in the candidate's browser. Candidates who already chose a human review keep it if you change the setting later, and links you have already sent keep working.

After submitting: transcripts and notes

When a candidate finishes, they can read the transcript of each of their answers and add a short note, for example to correct a word the transcript got wrong. Your team sees the note under that transcript in the interview report.

Practice question

You can turn on Practice question in Settings → Responsible AI. Before the first real video answer, candidates can record an answer to an easy warm-up question and play it back. It is never uploaded, stored or scored; it stays on the candidate's device. It is off by default.

Safeguards around scores

  • Contact details removed. Email addresses, phone numbers, web links and ID numbers are removed from transcripts before AI scoring.
  • Your rubric. Answers are scored against the criteria you set for each question.
  • Reviewer notes stay local unless you opt in. A reviewer's score correction changes only that candidate's score, unless an admin turns on Reviewer notes shape later scores. Then each affected score lists the notes that shaped it.
  • Calibration and drift. Scoring calibration & drift in Settings → Responsible AI shows how closely AI scores track your reviewers' scores, counting candidates who have both.
  • Humans decide. Scores support your team's judgement. Use adverse impact testing to check outcomes across groups.

Scoring calibration

In Settings → Responsible AI → Scoring calibration, an org admin can set:

  • Grading: Harsh, Balanced (default), or Generous. How strictly AI grades answers.
  • Score adjustment (−15 to +15): a fixed number of points added to or taken off every new AI score.

Select Save calibration. Changes apply to new scores and rescoring only. Existing scores don't change.

What you can see

Settings → Responsible AI also shows:

  • Active AI features: badges for the optional AI features running in your hiring process, such as Resume & application scoring, AI avatar interviews, AI voice phone screening, Culture-fit signal scoring, Response authenticity analysis, and Video response analysis.
  • Disclosure & opt-out activity (last 90 days): four counts: Disclosures shown, Continued with AI, Opted out, and Automated decisions. "Automated decisions" means a stage automation moved a candidate from an AI score alone, with no person clicking. Stage rules never reject on their own: they add the candidate to Ready to reject, and a person confirms. Your audit log (Settings → Security) shows which candidates and which rule.

AI decision trail per candidate

Every candidate profile has an AI decision trail card. Select Show to list up to the latest 50 events for that candidate, each with the AI feature involved:

  • AI notice shown
  • Continued with AI
  • Declined AI review
  • Asked for human-only review
  • AI step skipped (candidate declined)
  • Automated stage decision applied

The trail can't be edited. Use it to answer a candidate's question or a regulator's request about what AI did. Yearly counts of automated decisions by race and ethnicity, which Illinois requires, are in the AI-only decisions report.

Second-opinion grading panel

InsightHire also runs a panel of three independent AI graders plus a supervisor on interview answers. For now it runs in the background and is stored next to the normal score for comparison. It does not change any score or gate. See How AI interview scoring works for how the panel works, when it would become the score, and what is never automated.

Rules for AI in hiring differ by place and change often. The list of places above is our default, not legal advice. Check with counsel which notices, opt-outs and audits apply to your jobs.