Reading the transcript
The transcript is the raw record of the conversation. Here is how it's structured, what the timestamps mean, and how to use it alongside the AI evaluation.
Key takeaways
- The transcript shows every AI question and every candidate answer with timestamps.
- Use search (Cmd+F / Ctrl+F) to find specific words or topics quickly.
- Filler words (*um*, *uh*) are usually edited out for readability.
- If a phrase in the transcript looks wrong, listen to the audio at that timestamp — speech-to-text occasionally mis-hears.
Before you start
The transcript is the raw record of the interview conversation. Where the evaluation tells you what the AI thinks, the transcript shows you what was actually said. Knowing how to read it correctly is the single best skill for getting the most out of the platform.
How to open it
From the Interview Review page, click View Report in the header, or click any timestamp on the evaluation panel. The transcript view opens in a side panel or in a new tab depending on your screen width.
Structure
The transcript is organised as a sequence of turns. Each turn shows:
- Speaker label — AI Interviewer or Candidate.
- Timestamp — the moment the turn started, in mm:ss format.
- Body — the actual spoken text.
Long candidate answers are kept as single turns rather than broken up — the AI's listening logic uses 2-second silence pauses to decide an answer is complete. If a candidate paused mid-thought and the AI mistakenly moved on, you'll see that here.
Using search
Press Cmd+F (Mac) or Ctrl+F (Windows/Linux) to search within the transcript. Useful searches:
- Specific topics — search for kubernetes, team lead, budget to find every mention.
- Numbers and metrics — 73% or $1M — to verify quantified claims the candidate made.
- The AI's own language — search for Can you tell me about a time to find every behavioural-question lead-in.
Time-stamping audio playback
If you have the recording open in another tab, click any timestamp in the transcript to jump the playback to that moment. Useful when you want to check tone of voice for a specific answer — text strips away delivery.
Reading the transcript critically
TIP
The most common reading mistake is skimming for keywords (*kubernetes, leadership, scale*) and using their presence as a proxy for depth. Read the candidate's longest answers in full at least once — substance lives there, not in keyword frequency.
A few patterns to watch for:
- Direct answers vs deflections. A strong candidate answers the question asked. A weaker candidate pivots to a related but easier topic.
- Concrete examples vs abstractions. "In my last role, we did X, and it produced Y" lands much harder than "I believe strongly in X."
- Quantified outcomes. 73% cost reduction is verifiable; significant improvement is not. Senior candidates usually quantify.
- Awareness of own gaps. "I haven't directly managed people, but I led a 4-person matrix team for two quarters" is honest and useful; flat "I have lots of management experience" with no team size is a tell.
- Engagement with hard questions. When the AI asked a curveball, did the candidate reason out loud or stall?
When the transcript is wrong
Speech-to-text isn't perfect. If you see a phrase that doesn't make sense:
- Click the timestamp to jump the audio to that moment.
- Listen to the original. The AI is often almost right — Java heard as Jarvis, Kubernetes heard as cuber nettie, etc.
- Trust your ears over the text. Adjust your interpretation of the evaluation accordingly.
If you find repeated transcription errors that materially changed an evaluation, email the platform operator with the interview ID. Persistent errors on a domain (e.g., a specific technical vocabulary) help us improve the model.
Next: a deeper look at the evaluation report (if you came here directly).
Frequently asked questions
Is the transcript word-for-word accurate?
The transcript is generated by speech-to-text and is highly accurate for clear speech in supported accents. Mishearing happens occasionally — uncommon names, technical jargon, and heavy background noise are the usual culprits. If a phrase looks wrong, click the timestamp to jump to the audio and confirm.
Can I download the transcript?
Yes. On the Interview Review page, click **Download PDF** to get the full transcript + evaluation as a single document. For just the transcript, look for an export option in the transcript view's overflow menu.
Does the candidate see the transcript?
Not by default. Candidates can request their own transcript by emailing NextMantra (the email shown in the platform footer); we share it after the recruiter has reviewed the interview.