Mock interviews work. Here's the evidence.

The headline: coached candidates clearly do better

Across dozens of gold-standard trials — where people were randomly assigned to get coaching or not, so the comparison is fair — coached candidates consistently outperformed uncoached ones, and by a margin large enough to feel in a real conversation, not a rounding error. The gains showed up strongest in what people actually do and say — which is exactly what an interview measures.

And here's the part nobody expects: practice makes the interview more honest. In the field study closest to what we do, coached candidates' interview scores became twice as good at predicting how they'd actually perform on the job. Practice doesn't fake the signal — it removes the noise, so the interviewer finally sees the real you instead of your nerves.

The strongest proof: simulated interviewers changed real hiring

The most convincing evidence isn't about better scores in a lab — it's about who actually had a job months later. Six separate randomized trials had people practice with a simulated interviewer (a virtual character asking branching questions — the same idea as this product), then checked back on real employment:

Two details matter for you. The advantage was still there six months later — this is a skill you keep, not a warm-up glow. And the researchers traced the chain link by link: more completed practice sessions → better skills → more job offers. Measured, not assumed.

The catch: practice alone does almost nothing

Here's the finding that should change how you practice. In a controlled experiment, candidates who got extra self-practice time — recording themselves, re-recording, no one telling them anything — improved essentially not at all, in either performance or nerves. Repetition without a mirror is just repetition.

Add feedback and everything changes. Specific, per-answer feedback produces some of the largest improvements ever measured in the training literature — it raised graduates' interview confidence in dedicated studies and measurably lowered anxiety during mock interviews.

Why it works: the anxiety → confidence chain

Nervous candidates get lower interview ratings — yet those same nerves say almost nothing about how well they'll do the actual job. Read that again: anxiety makes good candidates look worse than they are. It's pure measurement error working against you — and it's trainable.

Practice attacks it from both sides. Mock interviews with feedback measurably reduced anxiety and raised confidence in student cohorts; and the "I've done this before" feeling is one of the strongest predictors of performing well under evaluation. Exposure plus visible proof you're improving is the classic anxiety treatment — applied to interviews.

The dose: how much practice is enough

• 3–5 sessions capture most of the benefit. Residency-interview outcome data found gains flattening after 3–5 mocks — and quality dropping past 6–8, when answers start sounding scripted.
• Space them out. Distributed practice beats cramming by 10–30% on retention — four sessions across ten days beats four in one evening.
• Completion matters. In the simulated-interviewer trials, the number of practice sessions people actually finished directly predicted their skills — which predicted offers.
• The night before: one light, confidence-priming run — not a marathon. Positive imagery reduced anxiety without over-rehearsal effects.

Simulated interviewers vs human mocks

The natural question: does practicing with an AI interviewer count? The strongest hiring outcomes above came from simulated interviewers — VR avatars with branching questions. Early studies of AI interview practice report improved confidence and articulation. What separates practice that works from practice that doesn't isn't who asks the questions — it's whether the session delivers immediate, specific feedback. A patient friend who says "that was fine" is a worse coach than a system that tells you your answer never addressed the question, your pace hit 190 wpm, and your voice read as anxious.

Two honest caveats from the research: most trials used small samples in specific populations, and head-to-head AI-vs-expert-human comparisons don't exist yet. What's established: simulated practice with feedback beats no practice by a wide margin, and beats unguided practice by design.

The evidence-based way to practice (what we'd do in your seat)

1. Pick the real target — a specific position, its posting, your CV. Generic practice transfers worse than role-specific practice.
2. 4 sessions across ~10 days — warm-up first ("tell me about yourself" is the highest-leverage single answer), then the phone screen, then the deep interviews.
3. Speak out loud, every time. Retrieval practice — actually producing the answer — beats re-reading your notes for retention.
4. Read every debrief and write down two concrete fixes per session (the deliberate- practice loop). Rerun 2–3 weak questions immediately while the correction is fresh.
5. Watch your numbers move — content score, fillers, pace, voice confidence — and stop at ~5 sessions per position. Sleep beats a sixth rehearsal.