When a vendor describes an interview copilot as DEI-focused, the phrase should mean three checkable things: the tool enforces structured interviews with the same questions and anchored rubrics for every candidate, it scores only the content of answers and never infers anything from face, voice, tone or accent, and it produces the records a bias audit needs. A tool that does those three things reduces bias because structure reduces bias; a tool that only adds a diversity tab to a dashboard does not. This page explains what the research supports, what the tool must never do, the main legal obligations in plain terms, and the checks a hiring team runs itself.
What structure does
Decades of selection research point the same way: interviews where every candidate gets the same questions, scored against behavioral anchors by interviewers who score independently before discussing, predict performance better and show smaller group differences than unstructured conversations. An interview copilot helps with the mechanics: it keeps the question set constant, prompts for the same probes, scores against the same anchors and records the evidence. The human side, independent scoring and committee discipline, is covered in how to reduce interview bias.
What the tool must never infer
| Input | Why it is a problem | Ask the vendor |
|---|---|---|
| Facial expression or eye contact | Correlates with culture, neurodiversity and camera setup, not competence | Confirm in writing that video is not analyzed |
| Voice tone, pace or pitch | Penalizes accents, speech differences and nervousness | Confirm scoring uses transcribed text only |
| Grammar and vocabulary as a proxy for clarity | Penalizes non-native speakers for roles that do not need native fluency | Ask how rubric anchors define clarity |
| Name, school, address or photo | Direct proxies for protected characteristics | Confirm these are excluded from scoring context |
| Gaps, dates or age signals | Age and caregiving proxies | Confirm the model sees the answer, not the résumé, when scoring |
CoPilot Interview's Interviewer Mode scores the content of transcribed answers against rubric dimensions and attaches the evidence it used; it does not analyze video, tone or facial expression. The candidate's view of these risks is in AI interview bias: what candidates should know.
Obligations in plain terms
- New York City Local Law 144: an automated tool that substantially assists hiring decisions for NYC jobs needs an independent bias audit within the previous year, a published summary of the results, and notice to candidates at least ten business days before use, with a way to request an alternative process.
- EU AI Act: AI used to evaluate or rank candidates is classed as high-risk. Employers deploying it must use it as instructed, keep logs, ensure human oversight, and inform affected workers; these obligations are scheduled to apply from 2026, a timetable the EU has been debating, so confirm the current dates with counsel.
- US states: Illinois requires notice, explanation and consent for AI analysis of video interviews, and Colorado's AI law adds impact-assessment and notice duties for consequential employment decisions. Several other states have bills in progress.
- Everywhere: existing anti-discrimination law applies to decisions made with AI assistance exactly as it applies to decisions made without it.
Candidates increasingly know these rights; AI interview laws and candidate rights is the page they read.
Checks you run yourself
- Compare pass-through rates by group at each stage, monthly, where collecting that data is lawful.
- Sample ten reports a month and read score against quote; disagreements of more than two points go to a calibration session.
- Rotate interviewers across candidate pools so no group is interviewed mostly by one person.
- Offer the alternative process in the notice, and track how often it is requested.
- Keep protected-characteristic information out of templates, notes and reports entirely.
Where CoPilot Interview fits
CoPilot Interview is best known as the AI interview assistant candidates use; Interviewer Mode is the hiring-side mode of the same desktop app, enforcing a fixed question set per role, scoring answer content against rubric anchors with evidence notes, and producing the per-interview records an audit needs, free on Windows and macOS with no signup. See Interviewer Mode. Fairness is mostly process; run the checks above with any tool, and follow your employer's rules and your counsel's advice on notice and audits.
Disclosure: CoPilot Interview is our product, so weigh what we say about it accordingly.
Frequently asked questions
Do AI interview copilots reduce bias?
They can, by enforcing structured interviews: same questions, anchored rubrics, independent scoring with evidence. A tool that infers anything from face, voice or accent does the opposite.
What should a DEI-focused interview tool never do?
Analyze facial expressions, eye contact, tone, pace or accent; treat grammar as competence; or include names, schools, photos or age signals in scoring context.
What does NYC Local Law 144 require?
An independent bias audit within the prior year, a published summary, candidate notice at least ten business days before use, and an alternative process on request, for automated tools substantially assisting NYC hiring decisions.
Is an interview copilot high-risk under the EU AI Act?
AI used to evaluate or rank candidates is in the high-risk category, which brings human-oversight, logging, notice and documentation duties for the deploying employer. Confirm the current application dates with counsel.
Does Interviewer Mode analyze video or tone?
No. It scores the content of transcribed answers against rubric dimensions and records the evidence it used.
Try Interviewer Mode on your next screen
Role-specific questions, live rubric scoring with evidence notes and a structured report at the end of the call. Free in the desktop app for Windows and macOS, no signup.
Get the desktop app