AI interview scoring against your rubric
Interview scores are only useful if they mean the same thing across candidates. Write the criteria your role needs, weight them, and every answer in every interview is graded against that one standard, with the transcript attached so your team can check the work.
Build the rubric once
Three decisions, made before the first invitation goes out, that every score afterwards refers back to.
Name the criteria
Write the things you actually hire on for this role, each with a short description of what a good answer looks like. Communication, technical depth, ownership, customer judgment: whatever the role needs.
Set the weights
Give each criterion a weight between 1 and 100. The set has to add up to 100, which forces the conversation about what matters most to happen before the interviews rather than after them.
Attach it to the interview
Mark one rubric as the organization default and override it per role. Every interview scheduled with that rubric is graded against it, so a candidate interviewed in March is measured the same way as one interviewed in June.
Weights are validated when the rubric is saved. A set that does not total 100 is rejected, which sounds pedantic until you have compared two candidates scored on rubrics that quietly disagreed about how much technical depth was worth.
What comes back from a scored interview
Scoring runs as the interview happens, answer by answer, and again at the end across the whole session. By the time a recruiter opens the report, it is finished.
- A score out of 100 for every individual answer, computed as a weighted average across your criteria.
- Two or three sentences of feedback per answer that name a specific criterion from your rubric rather than praising the candidate in general.
- An overall score and letter grade, from A plus down to F, for the interview as a whole.
- A short performance summary with top strengths and areas to improve, both tied back to the rubric.
- A readiness assessment for the role, so the report ends with a judgment rather than a table of numbers.
- Separate handling for quantitative reasoning questions, scored on method and accuracy instead of communication.
Every score shows its working
A number on its own is not reviewable. These four things are what make the score something a hiring manager can argue with.
The transcript sits next to the score
Each answer is shown in full alongside the score it earned, so a reviewer can disagree with the model and see exactly what it was reacting to.
Video playback is part of the report
The recorded answer plays inside the same report. Tone, hesitation, and the things a transcript cannot carry stay available to the person making the call.
Integrity scoring travels with the score
Every interview carries a 100-point integrity score. Tab switches, window blur, and fullscreen exits cost 20 points each, minor events cost 5, and every flag is timestamped against the question that was on screen.
Comparison puts candidates side by side
Pull the shortlist into one view with letter grades so the decision is made on like-for-like evidence instead of the interview everybody remembers best.
The same standard for everyone
The point of a rubric is not automation, it is consistency. A structured interview scored the same way every time is the single cheapest improvement most hiring processes can make, and it is the part that quietly falls apart when a busy panel runs the round.
- The scoring bands are calibrated so a candidate who genuinely answers the question lands in the 70 to 80 range, and scores below 50 are reserved for answers that miss the point. A model that scores everyone 90 tells you nothing.
- Candidate answers are passed to the model as data, with explicit instructions to ignore any directives inside them, so a candidate cannot write their way to a higher score.
- Every candidate for a role gets the same questions and the same rubric, which is the part of structured interviewing that human panels most often lose.
- Scores are a shortlist input, not a decision. The report is built to be reviewed by the people doing the hiring, with the evidence attached.
Interview scoring questions
How does AI interview scoring work?+
You define a scoring rubric of named criteria, each with a weight between 1 and 100 that together total 100 percent. As each answer comes in, the AI grades it against that rubric and returns a score out of 100 plus feedback that references at least one criterion by name. When the interview finishes, a final report computes an overall score and letter grade from the individual answers.
Can I use my own scoring criteria?+
Yes. Rubrics are written by your team, not chosen from a fixed list. Each criterion has a name, a weight, and a description of what a good answer looks like. You can keep several rubrics, set one as the organization default, and pick a different one per role.
What does an interview report contain?+
A per-question breakdown with the score and transcript of each answer, video playback of the recorded responses, an overall score and letter grade, a written summary with top strengths and areas to improve, a readiness assessment, and the integrity score from proctoring.
Can I compare candidates against each other?+
Yes. Shortlisted candidates for a role can be pulled into a side-by-side comparison with their scores converted to letter grades, so you are comparing performance against the same rubric rather than against your memory of each interview.
Does scoring work in other languages?+
Interviews run in 23 languages, and scoring runs on the answer as given. Reports are generated in the interview language so local hiring teams read them in the language they work in.
Is AI scoring fair to candidates?+
Scores come from the same rubric and the same questions for every candidate in a role, which removes the variation that comes from different interviewers on different days. The calibration is set so that genuine answers score in the 70 to 80 band rather than clustering at the top, and every score is published with the transcript behind it so a human reviewer can overrule it.
Related
AI interviews
The realistic avatar interviewer, proctoring, reports, and everything else in the platform.
Interview scheduling
Bulk invitations, deadlines, retakes, and live versus self-paced interviews.
Enterprise
Team roles, branding, credits, and organization-wide analytics.
Score your next shortlist the same way twice
Set up a rubric, invite a batch of candidates, and read scored reports with the evidence attached.