AI Interview

How an AI interview works

A typical AI interview for a first-round screen runs like this:

1. Setup. The recruiter defines the role, the competencies to assess, the questions, and what weak and strong answers sound like. Some tools draft questions from the job description for the recruiter to edit.
2. Invitation. Candidates receive a link by text or email and can take the interview any time, often within minutes of applying.
3. The conversation. The AI introduces itself as an AI, explains the process, and asks the questions. Unlike a one-way video interview, it responds to what the candidate says: if an answer is vague, it asks for a specific example; if the candidate asks about shift times, it answers from approved information.
4. Evaluation. The system transcribes the answers and maps them to the rubric, citing the parts of the transcript that support each rating.
5. Review. The recruiter sees a ranked list with summaries, scores and full transcripts or recordings, and decides who advances.

Voice-only AI interviews by phone or browser are common for contact-center and customer-facing roles, because the candidate's spoken communication is part of the job. Video and chat formats are also used.

Where AI interviews fit

AI interviews do the most good at the top of a high-volume funnel, where the alternative is that most applicants never get a conversation at all. A healthcare staffing firm receiving 1,500 applications a week for patient sitters and CNAs cannot phone-screen everyone in a timely way. With an AI interviewer, every applicant who passes basic requirements can be interviewed the same day, at 10 p.m. after a shift if that is when they are free.

They also help with consistency. Every candidate hears the same questions in the same way and is scored against the same rubric, which is the core of a structured interview. Tired interviewer effects and first-impression bias drop out.

They fit less well for final rounds, senior roles and situations where the candidate is evaluating you as much as you are evaluating them. AI interviews are a screening step, not a replacement for meeting the hiring manager. They are one part of the wider shift toward AI in recruitment.

What AI interviews should and should not score

Score what candidates say: the content of answers against job-related criteria, such as whether they gave a specific example, described sensible actions and reached a result. For roles where spoken language is a genuine requirement, scoring clarity or language proficiency can be job-related too, as long as it is measured at the level the role actually needs.

Do not score facial expressions, eye contact, perceived emotion, vocal tone as a personality signal, or appearance. The science linking these to job performance is weak, they are likely to disadvantage people with disabilities and some ethnic groups, and AI emotion recognition in workplaces and hiring is banned in the EU. Accent should never count against a candidate unless it genuinely prevents being understood in the role.

The safest design is evidence-based: every score links to quoted parts of the transcript, so a recruiter can check it and a candidate's challenge can be answered.

Setting up AI interviews fairly

• Tell candidates upfront that the interviewer is an AI, what it assesses, and how results are used.
• Use job-related questions and anchored rubrics, the same standard you would apply to a human-led structured interview.
• Keep humans responsible for decisions. Recruiters should review scores and transcripts, especially before rejecting anyone.
• Offer alternatives and accommodations. Candidates with speech, hearing or other disabilities, or poor connectivity, need a route to a human interview or an adjusted format.
• Test for adverse impact. Compare pass rates by sex, race/ethnicity and age group before launch and regularly after, and get an independent bias audit where required.
• Check validity. After a few months, see whether AI interview scores relate to later performance, quality scores and 90-day retention.
• Watch candidate feedback. Short surveys after the interview catch confusing questions and technical problems quickly.

Laws that apply

AI interviews are selection procedures, so Title VII, the ADA and the ADEA apply, and the employer is responsible for outcomes even when a vendor runs the tool. If the AI scores or ranks candidates for New York City jobs, it is likely an automated employment decision tool under Local Law 144, requiring a yearly independent bias audit, a published summary and advance notice. For recorded video interviews of Illinois applicants, the Illinois AI Video Interview Act requires notice, explanation, consent and deletion on request. Illinois also prohibits discriminatory AI use in employment decisions and requires notice. In the EU, AI used to evaluate candidates is high-risk under the AI Act, and GDPR restricts decisions made solely by automated means. Rules vary by state and country; confirm requirements with counsel.