Last updated 1 August 2026
A real conversation, not a recording
Most tools in this category ask a candidate to talk at a camera and analyse the recording afterwards. NorthAssay runs a live conversation instead — it hears an answer and asks the obvious next question.
The structural difference
Record-then-analyse is a form with a webcam. The candidate reads a prompt, records an answer, moves to the next prompt. Nothing that happens in answer one changes question two.
That design has a real consequence: the interview cannot follow up. A candidate gives a vague answer and the tool moves on, because it does not know the answer was vague until later.
| Record-then-analyse | Live conversation | |
|---|---|---|
| Follow-up questions | None — the script is fixed | Yes, based on what was said |
| Vague answer | Scored as given | Probed before moving on |
| Candidate experience | Talking at a camera | Being interviewed |
| Question order | Predetermined | Adapts to the conversation |
Why the follow-up is the whole point
The five checks that separate real skill from a good résumé are mostly follow-up questions. "What would you have done differently?" only works if it comes after a specific answer. "What broke?" only works if you heard the claim it is testing.
A fixed script cannot do that. It can ask the first question of the pair and never the second, which is why scripted video screens tend to surface how comfortable someone is on camera rather than how well they know the work.
What the candidate experiences
A browser tab, a microphone permission, and a conversation. No download, no scheduling, no waiting for a human interviewer to be free at 9pm in their timezone.
They know it is an AI interviewer. We do not dress it up as a person, and the interview is not a surprise — it is described before they start.
What this does not do
It does not analyse tone of voice, facial expression, or "enthusiasm". Those signals have a poor track record and a worse legal one, and we do not collect them.
It does not make a hiring decision. The interview produces a score with written rationale, the same as any other assessment type, and a person decides.
No bias audit has been conducted
We have limited exposure by design — the scorer never receives a name, photo or country — but limiting is not the same as measuring. Responsible AI sets out exactly what has and has not been tested.
See it interview you
The fastest way to judge this is to sit on the candidate side of it. Create a role, invite yourself, and answer three questions badly on purpose.
Try a live interview