How am I actually being scored - on what I said, or how I said it, or both?

Both, and they are reported separately rather than made into a single report.

On what you said. Your technical answers are scored with a per-question breakdown, so each answer is assessed on its own merit rather than blurred into one overall impression of how you did. On top of that you get a rating for every skill tested, along with the number of questions behind each rating - so a particular skill assessed across four answers is visibly better evidenced than one assessed on a single answer. The skills themselves come from the job description the interview was built from, which means you are measured against the role you applied for.

On how you said it. Communication is scored across four things: fluency, grammar, vocabulary and pronunciation. Alongside that, the conversation's overall sentiment, a sentiment timeline and emotion are tracked as context on the interview.

And on your interviewer's own judgement. They complete a structured scorecard with their rating and comments, which reaches the hiring team next to the AI report. The two are meant to be read together.

The scoring happens in the background and after the call, not live in front of you. An interview can be re-analysed, which regenerates the transcript, scores and summary - so a patchy recording doesn't have to be the final word. And how much of this applies depends on the plan the employer is on: technical and communication scoring and per-skill ratings come in above our free tier, which includes the live interview, AI notes, transcript, report, recording, interviewer scorecards and integrity checks. [confirm: exactly which scores are available on which plan]

If you are not sure which skills a round is assessing, ask at the start - knowing what is being tested is about being informed and not considered an undue advantage. To find out how you were assessed afterwards, you can always ask the recruiter when the decision has been conveyed to you. The scoring scale and range is proprietary and is based on a preset set of fixed rubrics that are customised according to the job role.