Every Question About AI Interview Software, Answered

Vettika runs 12-minute voice screening interviews. A candidate clicks a link, completes a 30-second mic check, then speaks with a voice AI that asks your custom questions. A real-time transcript and rubric-scored report land in your dashboard within seconds of the call ending.

Vettika delivers your defined question set in a natural voice and acknowledges responses conversationally. It does not improvise free-form follow-ups — that is deliberate. Bias reduction requires every candidate to receive the same questions in the same order. Free-form AI probing introduces unpredictable variation.

Runs on Gemini 2.5 Flash with structured rubric prompts. It evaluates for conceptual depth, answer specificity, and relevance — not keyword frequency. A technically correct answer delivered vaguely scores differently than a specific one with concrete examples.

Target length is 12 minutes, capped at 20 minutes to prevent runaway calls. Most candidates finish in 10–14 minutes depending on answer depth.

Long-winded: the AI redirects to the next question after a natural pause. Very short: it probes with a follow-up. Tangents: it acknowledges and bridges back. All three are deliberate behaviors, not edge cases.

A transcript and rubric scores appear in your dashboard within seconds. Any automation rules you configured — advance, flag for review, send feedback email — fire automatically based on score thresholds you defined.

Yes. Write questions manually, generate a set from a job description with one click, or mix both. The AI Campaign Builder generates a first-draft question set and rubric from a pasted job description in under 30 seconds. Every element is editable before you publish.

Yes. No credit card required. You get 3 interviews that never expire. Use them on real candidates or try the experience yourself by clicking your own apply link.

One call where the candidate audibly spoke and the AI asked at least one question. If they no-show or drop before saying a word, it doesn't count against any cap — paid or free.

Per interview conducted. Not per seat, not per job posting. You only pay when a candidate actually completes an interview.

You're prompted to buy a pack (10 for $19 or 50 for $79) or pick a monthly plan. Your existing interviews and transcripts stay accessible regardless.

Yes — 12 months from purchase. Use them whenever you need them inside that window.

You're not cut off. Overage interviews are billed at the per-interview rate for your tier ($1.50 on Pro, $1.00 on Agency). We don't block candidate access — a failed interview mid-hire is worse than an overage charge.

Yes — 20% off Pro and Agency when you pay annually. Packs are one-off purchases and are not affected by the billing toggle.

HireVue is enterprise-only with five-figure annual contracts. Karat uses human interviewers at $150–$500 per interview. Vettika uses AI scoring at a fraction of that cost, with no minimum commitment.

Yes. Pack credits draw down first; the monthly cap is only consumed after pack credits exhaust.

In most jurisdictions yes, if implemented correctly. The most prominent regulation is NYC Local Law 144, effective July 2023, which requires an annual bias audit and candidate notification before any AEDT is used. Vettika publishes its bias audit publicly at /legal/aedt-bias-audit and surfaces required disclosures on every apply link automatically — you don't need to configure anything.

Local Law 144 applies to employers and staffing agencies that use automated employment decision tools (AEDTs) to screen candidates for NYC jobs. It requires: (1) an annual bias audit by an independent auditor, (2) public posting of results, and (3) candidate notification before use. Vettika satisfies all three.

The design intent is the opposite. Every candidate receives the same questions in the same order delivered by the same voice — no interviewer fatigue, no affinity bias, no gut-feel culture-fit judgment in round one. The annual independent bias audit verifies outcome scores don't correlate with protected characteristics.

Yes. Every apply link displays a disclosure banner before the interview begins. The banner includes a link to the accommodation request pathway at /legal/accommodations. A recruiter reviews accommodation requests and can offer an alternative screening format.

No. Vettika outputs scores and a recommended action based on your rubric. All hiring decisions are made by a human recruiter. This is a legal requirement under most AEDT regulations. The platform is explicitly designed as a decision-support tool, not a decision-maker.

Similar to leaving a detailed voicemail or doing a phone screen — but without the scheduling friction. No video, no waiting room, no interviewer running late. Candidates open the link, check their mic, and speak. The voice is natural, the questions are clear, and the experience typically takes 10–14 minutes.

No. The entire interview runs in the browser. No app, no Zoom, no account required. Any modern smartphone or desktop browser works.

Scores are private to recruiters by default. You can configure an automation rule to send an optional feedback email post-decision that shares selected insights without exposing raw rubric scores.

Yes. The disclosure banner on every apply link includes an accommodation request pathway before the interview starts. Candidates who need an alternative format can request one, and a human reviews the request. See /legal/accommodations for the full policy.

Native two-way integrations: Greenhouse (Harvest API), Lever, Workable, Airtable, and Google Sheets. A generic webhook adapter handles any other ATS or HRIS. Candidate data syncs both directions — status updates in Vettika reflect back to your ATS.

Candidate data is stored in managed Postgres hosted in US-East (Neon). Audio is processed in real-time and is not retained beyond the session. Transcripts and rubric scores are stored and accessible to the recruiter who owns the campaign.

No. Candidate interview data is never used to train foundation models. The underlying voice AI uses Deepgram for speech-to-text and Google Gemini for evaluation — both operate under data processing agreements that prohibit training on customer data.

Yes. Email support@vettika.com with a deletion request. Data is removed within 30 days and includes the transcript, rubric scores, and associated metadata. The recruiter is notified when deletion completes.