How answers are scored
Each of your answers is scored on relevance, depth and clarity, each from 0 to 100, and gets a label of Strong, Promising, Needs practice. The session gets an overall score from your own answers. Delivery feedback is speech-based: speaking pace, filler words (in English answers) and pauses, measured from your spoken answer. When there is too little of your own to score, the report says why instead of showing a number.
Explanation · Last checked against the product on
What each score measures
The AI scores every answer you gave on relevance, depth and clarity, each from 0 to 100. Relevance is whether you answered the question asked. Depth is how much evidence and detail you gave. The third is how easy the answer was to follow. On the Responses tab each answer shows its score out of 100 and one of these labels:
- Strong
- Promising
- Needs practice
The overall score at the top of the Overview covers only the answers you gave. If a session ended early, the report says the score covers only the answers you reached. The rubric behind each score, with worked examples, is on the methodology page.
Delivery
Delivery feedback is speech-based: speaking pace, filler words (in English answers) and pauses, measured from your spoken answer. There is no facial-expression, eye-contact or posture analysis: the camera is only your self-view. The How you said it panel at the top of the Responses tab shows:
| Measure | What it counts |
|---|---|
| Pace | Your speaking speed, read as slow, comfortable or fast. Spoken answers only. |
| Long pauses | Silences in the middle of an answer, with the longest one. Spoken answers only. |
| Filler words | Words such as “um” and “like”, counted in English answers. |
| Hedges | Phrases such as “maybe” or “I think” that soften what you did. |
| “I” vs “we” | In behavioural answers, how often you said what you did yourself. |
| Average answer | Answer length in words, with the range this format usually lands in. |
Everyone sees the session totals. The per-answer breakdown is included with Pro. A session you typed has no pace or pause figures.
When a report has no score
A report shows no overall number, and says why in one sentence, when:
- your answers are still being scored;
- the AI answered every question on your behalf;
- no usable answers were captured;
- too few of your replies took on the question asked;
- the session was cut short by a problem on our side.
Your answers and their feedback stay on the Responses tab either way. If scoring failed, use Retry analysis.
Answers the AI wrote for you
If you used Let AI answer during a session, that question is marked Answered by AI on the report. It is shown as a model answer and is not scored. It never counts toward your score or your overall.
The Overview breaks those answers down under The AI’s answers, explained: what an interviewer would credit, moves you can reuse and what to replace with your own experience. The percentages there rate the AI’s answers as worked examples. They are not your score.
Scores in other formats
A visa or citizenship practice leads with the officer’s decision, and its number is labelled Practice score. A role-play or negotiation is scored on your goals and how you ran the conversation, so each turn is judged as a move. See Report formats.