Eight questions for your job, asked one at a time by an AI interviewer who reads them aloud, starts a clock and follows up on what you actually said. You answer out loud, the way you will on the day. At the end, one report scores every answer and shows what to fix first.
Questions for 190 jobs come from our question bank. Any other job gets the common questions every interviewer asks.
The questions come from our company pages, written for that company and job. Nothing here guesses a company’s hiring steps.
With the AI on, your resume text goes with each answer to our server and Google Gemini so the follow-ups can ask about your real work. A file is read in this browser; the text of a resume you upload is saved to ClapAssist (see or delete).
Your answers are recorded in this browser. Each one goes through our server to Google Gemini for the transcript and the review. Nothing is kept on our servers.
Four steps, about half an hour, nothing to install.
Most people do well alone and freeze in the room. A simulator puts the pressure back.
You hear the question, you do not read it. The interviewer speaks each question, so you answer the way you will on a call, with no time to polish a sentence in your head.
There is a clock. Each answer gets 1, 2 or 3 minutes and stops by itself, so you learn what a two-minute story feels like.
It listens and digs. After you answer, it asks about the vaguest part of what you said: how you knew it worked, what you did yourself, what happened next. That is the moment most practice skips.
It keeps going. Eight questions in a row, mixed: why this job, stories from your work, what you would do, and questions about the job itself. Stamina is part of the test.
It counts, it does not guess. Your pace, filler words, pauses and hedges are counted by fixed rules from a word-for-word transcript, so “um” is never tidied away.
Made by this page’s own engine, the same one that counts your answers. Type “Nurse” in the job box and a full interview can look like this.
One run of the draw. Press “Different questions” for another set; questions for this job come from 30 written for it, mixed with the ones every interviewer asks.
The question: Tell me about a time you noticed a patient getting worse before the numbers made it obvious. What did you see, and what did you do?
Um, so on a night shift last year I had a patient two days after surgery who was, uh, more restless than on my last round. His numbers still looked okay, but I think his breathing was a bit faster and he kept pulling at his gown. I did a full set of observations, and his oxygen had started to dip, so I sat him up and called the doctor with SBAR. Basically I said what I saw, his numbers, and that I was worried. They came within ten minutes and he was, like, treated early for a chest infection. Now I always trust that restless feeling and check again.
| What the report counted (47 seconds, 110 words) | Result |
|---|---|
| Pace | 140 words a minute, easy to follow |
| Filler words: “basically” ×1, “like” ×1, “uh” ×1, “um” ×1 | 4 (3.6% of words) |
| Hedges and weak words: “I think” ×1 | 1 |
| Numbers said: “two”, “ten” | 2 |
| Length | 0:47, about right |
| Delivery: pace 25/25, filler words 23/25, length 15/15, hedges 10/10, repeats 10/10. Pauses are left out here: they need the microphone level of a real recording. | 97/100 |
In the real report the AI coach adds a content mark out of 10, which parts of the story it heard (situation, task, action, result), the fix that matters most and your answer rewritten from your own facts. Then the interviewer asks a follow-up, often one of these: What would you have done if the doctor had said to keep watching him? Which of the early signs do you think nurses miss most often?
Two parts. Delivery is counted by fixed rules on your words and the microphone level. Content is the AI coach’s mark out of 10, with five named parts. Practice mode hides both.
| Delivery, out of 100: what it checks | Points |
|---|---|
| Pace: full marks from 100 to 170 words a minute, none at 50 outside that band. Microsoft’s Speaker Coach suggests 100 to 165; Yoodli calls 130 to 170 typical. | 25 |
| Filler words (um, uh, er, like, you know, I mean, basically…): full marks under 3% of your words (Yoodli’s goal), none at 10%. | 25 |
| Pauses: a pause to think is fine. Each pause of 3 seconds or more costs 5, and so does starting after more than 5 seconds. | 15 |
| Length: full marks from 30 seconds to 2 minutes (3 for behavioral and technical questions), from Indeed’s advice of 30 seconds to 4 minutes. | 15 |
| Hedges and weak words (I think, maybe, probably, really, just): full marks at 2 per 100 words or fewer, none at 8. | 10 |
| Repeats: words said twice in a row, phrases said 3 or more times, and one opener starting a third of your sentences cost 2 each. | 10 |
Content, out of 10 (the AI coach): 1 to 2 no real answer; 3 to 4 answers the question but vague; 5 to 6 a real example but thin on what you did or the result; 7 to 8 a clear example, your own actions and a result; 9 to 10 specific, measurable and tied to the job.
The five parts of the content, 1 to 5 each, shown after every answer with how to raise the low ones: Relevance (it answers the question asked), Structure (a clear order), Specificity (real names, numbers and moments), Impact (what changed because of you) and Clarity (short and easy to follow). A part the coach did not rate says “not rated”; it never counts as a zero.
Overall: content 60%, delivery 40%. 80 or more is strong, 60 to 79 is good, under 60 needs work. A typed answer has no delivery to measure (no pace, pauses or fillers), so an interview you type is marked on its content alone, and without the AI coach it gets no mark at all rather than a guessed one.
A practice interview that behaves like a real one: an interviewer asks one question at a time, you answer out loud without notes, and you get told how it went. This one runs in your browser. The AI interviewer reads each question aloud, times your answer, asks a follow-up on the weakest part of it, and ends with a report on every answer.
Yes, with no account and no paywall at the report. The AI review has a daily allowance per connection of 40 answers, which covers a full simulated interview with all its follow-ups. Without the AI you still get the counts: pace, filler words, pauses, length, hedges and repeats.
It asks 8 questions, and the interviewer follows up on what you said: up to two follow-ups a question with the default Neutral interviewer (one with Friendly, three with Tough), stopping sooner when your answer is already specific. Each answer has a 2-minute clock by default (you can pick 1 or 3), and most answers take less, so set aside thirty to forty-five minutes. You can end at any time and still get the report for the answers you gave.
Type your job and the questions come from our bank of practice questions for 190 jobs, from nurse and teacher to data analyst and software developer, mixed with the questions every interviewer asks. For any other job, or a specific opening, paste the job post and the AI writes the questions from it.
Yes, in the phone’s browser. Allow the microphone when asked. Phones do not show live captions while you speak; your words appear after each answer, when the AI writes them down. If you are somewhere you cannot talk, choose Type my answers.
Your answer is recorded in this browser tab. When you stop, it goes through our server to Google Gemini to be written down word for word and reviewed, and nothing is kept on our servers. The recordings are gone when you close the tab; only the scores of past interviews stay, in this browser, with a delete button.
Yes. Pick the Tough interviewer and it follows up twice on the same answer, the way a panel digs into a story. Set Your experience to Senior for questions meant for experienced people, or Behavioral or Technical to keep to one kind of question.
The interview is kept in this tab. A reload brings you back to the question you were on, with your earlier answers and their reviews; the browser’s Back button goes one screen back, not off the page. Only an answer you were recording at that moment is lost.
It is the same app. This page opens ready for a whole interview: 8 mixed questions, read aloud, with follow-ups, in one sitting. The AI mock interview page is set up for shorter rounds, one kind of question at a time, or a list of your own questions.
The real interviewer will ask something you did not practice. ClapAssist is an AI interview assistant for Mac and Windows that listens to the call and shows you what to say, in the same plain style you practiced here.
Download with 10 free minutes