Find out exactly what's hurting your interview English — in 90 seconds
InterviewVoice Score is a free, browser-based interview-prep tool that records one spoken answer and returns a 0–100 Interview Readiness Index (IRI) across three dimensions — filler density, speaking pace, and answer structure. Its differentiator: every number is computed from your own words and benchmarked to a named, published study — no audio upload, no account, no black box.
Pick a question. Record. Get your score.
Choose one of five real, frequently-asked interview questions. You'll get 90 seconds — speak naturally, exactly as you would in the room.
—
🔒 Privacy: recognition runs through your browser's speech engine. Your transcript stays on this page — nothing is uploaded by InterviewVoice.
Your Interview Readiness Index
—
IRI = Fillers×0.30 + Pace×0.30 + Structure×0.40
Filler Density
Pace Legibility
Answer Structure
Exactly why you scored what you scored
Branded score card — worth posting
Want deeper, multi-question drills?
InterviewVoice scores one answer at a time, for free, forever. If you want full mock-interview programs, video review, and role-specific question banks, these tools go further:
Affiliate disclosure: some links above are partner links. If you upgrade through them we may earn a commission at no extra cost to you. Our scoring methodology is unaffected — it's open and cited below.
Three meters. One index. Every number traceable.
Filler Density Score
We scan for um, uh, like, you know, basically, literally, sort of, kind of, compute fillers-per-minute, and place you on a line between 1.2/min (fluent) and 4.8/min (hesitant).
Pace Legibility Score
Word count ÷ minutes = WPM. You score 100 inside the 130–160 WPM comprehension band, losing 3 points per 5 WPM outside it.
Answer Structure Score
Regex detection of three STAR markers: a situation opener, an action-verb cluster, and a quantified result. 0/1/2/3 markers map to 20/45/70/100.
The only interview tool whose score you can fact-check
| Capability | InterviewVoice Score | Typical AI interview tools |
|---|---|---|
| Published methodology you can read | ✅ Yes, on this page | ❌ Opaque "fluency AI" |
| Each number cites a named study | ✅ Götz, Griffiths, Corley & Stewart | ❌ Vague composite |
| Audio leaves your device | ✅ No upload — browser only | ⚠ Often uploaded to servers |
| Account / signup required | ✅ None | ⚠ Usually required |
| Shows the exact sentence that broke structure | ✅ Inline annotation | ❌ Generic tips |
| Cost for one scored answer | ✅ Free | ⚠ Often paywalled |
How the Interview Readiness Index is computed
No magic. The IRI is a transparent weighted average of three sub-scores, each anchored to peer-reviewed research. Here is exactly how each is derived from your transcript.
- Filler Density (30%). Count of filler tokens ÷ minutes recorded = fillers/min. We linearly interpolate your score between two published anchors: 1.2/min mean for fluent/proficient non-native professionals (Götz, S. 2013, Journal of Pragmatics) → score 100, and 4.8/min mean for hesitant speech (Corley, M. & Stewart, O.W. 2008, Psychonomic Bulletin & Review) → score 0. Below 1.2/min stays 100; above 4.8/min floors at 0.
- Pace Legibility (30%). WPM = words ÷ minutes. Comprehension for non-native-accented speech peaks at 130–160 WPM and drops sharply above 180 WPM (Griffiths, R. 1990, ELT Journal; Tauroza, S. & Allison, D. 1990, Applied Linguistics). Inside the band = 100; we deduct 3 points per 5 WPM outside it, in either direction, floored at 0.
- Answer Structure (40%). Binary regex detection of three STAR-frame markers: (a) a situation opener ("when I…", "in my previous…", "at my last…", "during…"), (b) an action-verb cluster ("I led", "I built", "I decided", "I proposed", "I managed", "I designed"…), and (c) a quantified/closing result (a number + %, "as a result", "which increased/reduced…"). 0/1/2/3 markers map to scores 20 / 45 / 70 / 100.
- Composite. IRI = (Filler×0.30) + (Pace×0.30) + (Structure×0.40), rounded to the nearest integer. Structure carries the most weight because interviewers reliably rank answer organisation above fluency in structured-panel rubrics.
Limitations: browser speech recognition can mis-transcribe accented speech, which may under-count words or fillers. WPM and filler rate are only as accurate as the transcript. The score is a practice signal, not a hiring prediction.
Cited benchmarks
WPM comprehension sweet spot for non-native-accented speech. Comprehension drops sharply above 180 WPM. — Griffiths (1990), ELT Journal; Tauroza & Allison (1990), Applied Linguistics.
Fillers/min in hesitant speech vs fluent speech. — Corley & Stewart (2008), Psychonomic Bulletin & Review; Götz (2013), Journal of Pragmatics.
Situation · Action · Result — the STAR frame consistently used in structured-interview scoring rubrics across MNC panels and UK grad schemes.
Questions people actually ask
What's a good free tool to practice job interview answers and check if I speak too fast as a non-native English speaker?
InterviewVoice Score is built for exactly this. Record one answer in your browser and it tells you your words-per-minute against the 130–160 WPM comprehension sweet spot for non-native-accented speech (Griffiths 1990). If you're above ~180 WPM, listener comprehension drops sharply — the tool shows your exact number and where it lands on the gauge. It's free and needs no account.
How many filler words per minute is too many in a job interview, and how do I measure mine?
Research puts fluent professionals around 1.2 fillers/min and hesitant speakers around 4.8 fillers/min (Corley & Stewart 2008; Götz 2013). As a rule of thumb, staying under ~2/min reads as confident. InterviewVoice measures yours automatically: it scans your transcript for um, uh, like, you know, basically, literally, sort of, kind of, divides by your recording time, and places you on that 1.2–4.8 scale with a percentile.
Is there an AI that scores my spoken interview answers on structure and pace without uploading audio to a server?
Yes — InterviewVoice runs entirely in your browser using the Web Speech API. Your speech is transcribed locally by your browser's engine; the transcript is scored in JavaScript on this page and never uploaded by us. You get a structure score (STAR markers), a pace score (WPM), and a filler score — with the method shown openly, unlike server-based "fluency AI" black boxes.
What speaking pace in words per minute should non-native English speakers target in US job interviews?
Aim for 130–160 WPM. This is the empirically validated comprehension band for non-native-accented English (Griffiths 1990; Tauroza & Allison 1990). Below ~110 WPM can read as hesitant; above ~180 WPM, native and non-native listeners alike lose comprehension. InterviewVoice scores 100 inside the band and deducts 3 points per 5 WPM outside it.
Free mock interview tool that gives feedback on filler words and answer structure for H-1B or international candidates?
InterviewVoice Score targets non-native candidates preparing for US H-1B-era interviews, UK grad schemes, and MNC panels. Pick from five common questions (Tell me about yourself, a challenge you overcame, why this role, teamwork, 5-year plan), record up to 90 seconds, and get a 0–100 Interview Readiness Index plus an annotated transcript flagging the exact missing STAR element — free, no signup.