Practice out loud, on camera and change one thing between takes. Rehearsing answers in your head misses what a recording shows: how fast you speak, which filler words you lean on and where you go quiet.
The routine
1.Pick a role and one question
Choose the role you are interviewing for. Each role has 15 questions in the tool: role-specific ones and general ones such as “Tell me about yourself”. Start with the one you dread most. Read the role guides.2.Record one take
Answer as if it were the real interview, without notes. The recording stops at 3 minutes on the trial and 5 minutes on a paid plan. Do not stop to restart: the first take shows your habits.3.Read the numbers before the text
Look at pace, filler words and long pauses first. They are specific and they are hard to notice while you are talking. Then read the written feedback.4.Change one thing
Pick the single biggest problem: the answer has no result, you rush, you fill every gap with the same word. Fixing one thing lets you see whether the change worked. Fixing four teaches you nothing.5.Record the same question again
Compare the same three numbers with the first take. If the problem moved, keep the change and pick the next one. If it did not, try a different fix.
Every account gets 2 trial answers a month. A 6-answer pack has 6 and Premium allows up to 60 a month. With only a few takes, use the first to find the problem and the next to test one fix.
The three delivery numbers
| Number | How the tool measures it | What to practice |
|---|---|---|
| Pace | Words in your answer divided by its length in minutes. The tool treats 100 to 160 words per minute as in range and highlights anything outside it. | Above the range, shorten your sentences and pause between them. Below it, cut the lead-in and move to the point sooner. |
| Filler words | Counts "um", "uh", "you know" wherever they occur. "like", "so", "right", "actually", "basically", "literally", "honestly" are ordinary words too, so they count only when the transcript puts a comma after them, as in “So, I led…”. | Replace the word you use most with a silent beat. Look at which fillers the report names and fix the top one before the others. |
| Long pauses | A silence of 2 seconds or more between two words. | A short pause to think is fine. A long one can mean you lost the thread, so write down the three points of your answer and rehearse the order. |
The ranges are the tool's own settings. They are a sensible reference, not a rule that interviewers apply.
A weak answer and a strong one
The question: Walk me through a technically challenging project you've owned end to end. The two answers below are written as they would be spoken.
Weak
So, um, I worked on this project where we were basically rebuilding part of our backend and it was pretty challenging because there was a lot of legacy code and I had to, like, figure out how everything worked before I could really start changing anything, so that took a while and then once I understood it I started making changes and eventually we shipped it and it worked pretty well I think.
No system design vocabulary, no trade-off named, no metric: one 45-second run-on sentence.
Strong
I owned the migration of our payments service off a monolith. The hard trade-off was consistency versus availability during the cutover: I chose a dual-write pattern with a reconciliation job rather than a hard cutover, because a failed payment is worse than a slow one. We caught two data-drift bugs in staging that the reconciliation job surfaced before they hit production. Migration completed with zero customer-facing incidents.
Four sentences, each carrying one idea. "Trade-off," "dual-write," and "reconciliation" are the system-design vocabulary interviewers listen for: say each one once, don't circle back to repeat it for emphasis. Try the half-beat pause after "cutover" before the metric; most candidates rush straight through it.
Role guides, each built around its own job, are available for Software Engineer, Product Manager, Data Analyst, Nurse, Teacher and Recent Graduate.
Practice the way you will be judged
In our study of employer policies, 13 of the 14 that state a rule for live interviews prohibit AI. Preparation is allowed almost everywhere, but the interview itself is unaided. So rehearse the take without notes and without an assistant open, even if you used both to prepare. What employers allow, stage by stage.
Four ways practice goes wrong
Practicing only in your head
You cannot hear your own fillers or feel yourself speed up. A recording can.Changing everything between takes
You will not know which change helped. Change one thing, then compare.Stopping at the score
A number tells you where you are. The next take, with one fix, is what moves it.Reading from a script
A script fixes the words and hides the delivery problems you came to find. Use a short outline of the points, in order.
Common questions
How many times should I practice one question?
Until the three delivery numbers stop changing much between takes and you are no longer fixing the same problem. Two or three takes of one question teach more than one take of five questions, because each take shows whether the last change worked.
How long should a practice answer be?
The tool stops a recording at 3 minutes on the trial and 5 minutes on a paid plan. Those are limits, not targets. Aim to finish the answer, with a result in it, well inside them.
Is my video stored?
No. Your audio is sent for transcription and discarded afterwards and recordings of your video or audio are never stored. The full policy is on the privacy page. Read the privacy policy.
Can I practice with notes or an AI open?
You can use both to prepare, but practice the take itself without them. Of the 14 employers that state a rule for live interviews, 13 prohibit AI, so the condition to rehearse is an answer given unaided. See what employers allow.
Keep going
How to read a mock interview feedback report
What each part of a report means and what it cannot tell you.
Practicing for an interview with AI
What employers allow at each stage, from their own published rules.