Product · Candidate experience

Candidate experience is the whole product

A first-round interview is most candidates' first real interaction with your company. In 2026 it is increasingly also their first interaction with an AI — 63% of candidates have now done an AI interview (Greenhouse, 2026 Candidate AI Interview Report, April 2026, N=2,950). The same report explains what's at stake when it goes wrong: 38% of US candidates have already withdrawn from a hiring process because it included an AI interview. The top reasons aren't "it was AI" — they're one-way formats with no human present (33%) and undisclosed AI use (27%).

So we treat candidate experience as a set of design commitments you can check, not a slogan. Here they are.

Our five commitments

1. Candidates always know they're talking to an AI

Disclosure happens before the interview starts, in plain language: this is an AI interviewer, it evaluates only what you say, and your recruiter reads the full transcript. No simulated humanity, no afterward-surprise. (70% of candidates report not being clearly told upfront that AI would evaluate them — that industry default is the trust problem, and it's also increasingly a legal one: Illinois' AIVIA consent rules tightened in January 2026.)

2. It's a real two-way conversation

A live voice interview that asks your structured questions, listens, and follows up on what the candidate actually said. Candidates can think out loud, recover from a weak first sentence, and ask questions back. The 33% who walk away from pre-recorded, no-human-present video aren't wrong about the format — talking at a camera isn't an interview. We don't offer one-way video at all, by design.

3. The transcript is the record — and it's visible

Every interview produces a full transcript your recruiter sees. The scored report cites the rubric you defined, against what the candidate said. When a candidate asks "how was I evaluated?", there is a real answer.

4. Content-only evaluation

Scoring is content-only — no tone, facial, or biometric analysis. We evaluate what was said, never how it sounded or looked. Answers are scored against the rubric you define, and an automated guardrail in our build pipeline blocks emotion-, facial-, or prosody-analysis dependencies from ever entering the codebase. This isn't only an ethics position; it's what makes scores explainable to the hiring manager and defensible if challenged.

5. A human makes the decision

Vettika conducts the structured first conversation and delivers evidence. Advancing or rejecting a candidate is your recruiter's call, made with the transcript in front of them.

How we'll prove it (and why there are no numbers on this page yet)

The AI-interview category has a measurement problem: vendor-reported candidate-satisfaction numbers with no published methodology. We've made the opposite bet — we don't publish candidate-experience metrics until we can show our work.

Instrumentation for two numbers is being built now:

Candidate NPS, asked at the end of every interview (one 0–10 question, anonymous, aggregated)
Completion rate — interviews started vs finished, by campaign

When they go live, this page will carry both numbers, updated on a stated cadence, with the methodology (sample, period, question wording) published alongside. Measured across real interviews, not a case study we picked. Until then: the commitments above are checkable today — run one interview and read the transcript yourself.

How the incumbents compare

We maintain evidence-based, date-stamped comparisons — including where each competitor is the better choice:

Run one and judge it as a candidate would

The honest test of any of this is doing an interview yourself. Create a campaign, take your own first-round, read your transcript. vettika.com.

Run your first 3 free See pricing