Reviewing Results
Learn how to review AI interview results, understand scores, and use Ezra's evaluation tools including communication analysis and cheat detection.
This page covers results from AI interviews — screening interviews conducted by Ezra. If your company uses live interviews, those have a separate review experience.
Candidate list
Your candidate list shows all candidates for a job in a table with their score, name, email, job title, status, and interview date.
Candidate statuses include:
- Unregistered — The candidate has been invited but hasn't opened the interview link yet
- Registered — The candidate has opened the interview link and registered
- In Progress — The candidate is currently taking the interview
- Completed — The candidate has finished the interview
- Incomplete — The candidate started the interview but did not finish or did not provide sufficient answers
- Opted Out — The candidate declined to participate
Clicking a candidate's row to view their evaluation will automatically mark them as reviewed. You can also click the eye icon to toggle their reviewed status. Use the kebab menu (three dots) to archive, restore, or mark a candidate as opted out.
Evaluation tabs
Clicking a candidate opens a detail panel with several tabs. The tabs you see depend on your company's settings and what data is available.
Summary
The summary tab gives you a quick overview of the candidate's performance:
- Interview questions — Each question is listed with its score (1–4) and a View details link to see the full answer. Click the play button to jump directly to that answer in the video. Play buttons are also available throughout the transcript, so you can jump to any moment in the recording.
- Strengths — Key positives Ezra identified based on the candidate's interview and resume.
- Areas for growth — Concerns or gaps Ezra flagged.
- Communication scores (if enabled) — A compact overview of the candidate's communication performance across six dimensions.
- Integrity assessment (if enabled) — Meters showing cheat detection, candidate fraud warning, and resume authenticity levels.
Transcript
The full interview transcript with speaker labels (Ezra / Candidate) and timestamps. Linked to video playback so you can click any segment to watch the corresponding moment.
Relevant info
Key insights Ezra detected during the interview, including additional notes and any questions the candidate asked.
Suggested follow-ups
AI-generated questions recommended for the candidate's next interview round, based on their performance and any areas that warrant deeper exploration.
Communication analysis
If communication analysis is enabled, this tab shows an overall communication score (0–100%) broken down into six dimensions:
| Dimension | What it measures |
|---|---|
| Clarity | How clearly the candidate expressed their thoughts |
| Fluency | Smoothness and flow of speech |
| Structure | Organization and logical flow of responses |
| Confidence | Self-assurance and conviction in responses |
| Engagement | Level of interaction and enthusiasm |
| Vocabulary | Range and appropriateness of word choice |
Score ranges: Excellent (80–100%), Good (60–79%), Below Average (40–59%), Needs Improvement (0–39%).
Communication analysis does not affect the overall interview score by default. It is an additional reference point to help you understand how a candidate communicates, independent of the content of their answers. If you want communication quality to factor into the score, you can set a weight (0–20%) in interview settings — only then will the communication score be blended into the candidate's overall score by the percentage you chose.
Cheat detection
If cheat detection is enabled, Ezra assesses the likelihood that a candidate was reading from a script or using external tools during the interview.
The assessment uses a five-level scale: Very Low → Low → Medium → High → Very High.
Cheat detection does not affect the overall interview score. It is a separate signal to help you identify which interviews may need a closer look. You can also manually override the cheat detection level on any candidate to reflect your own assessment after reviewing the recording.
Ezra utilizes academic research in linguistics, natural language processing, and speech analysis to make its assessment. Some patterns it looks for are:
- Prosodic patterns: unnatural rhythm, stress, and intonation
- Temporal markers: extended pauses, reading-like cadence
- Phonetic consistency: lack of natural speech variation
- Acoustic artifacts: keyboard sounds, paper rustling
- Disfluency patterns: absence of natural hesitations and repairs
Fraud detection
If fraud detection is enabled, this tab provides two assessments:
- Candidate fraud warning (Low / Moderate / High / Critical) — Checks signals such as email account age, phone number type, location consistency, and experience and education verification against third-party data.
- Resume authenticity (Low / Moderate / High / Critical) — Analyzes the resume for suspicious patterns including unlikely career trajectories, achievement patterns, and timeline gaps.
Fraud detection does not affect the overall interview score. Like cheat detection, these are additional reference points to help you determine which candidates may warrant further verification.
Scoring
Each interview question is scored on a 1–4 scale:
| Score | Label |
|---|---|
| 1 | Poor |
| 2 | Fair |
| 3 | Good |
| 4 | Excellent |
Scores are assigned automatically by Ezra based on the benchmarks you set during training. You can manually override any score by clicking on it and entering a new value.
The average score is displayed on the candidate list with color coding so you can quickly identify top performers.
Export results
You can export interview data as a CSV file from the job's interview data page. Choose which columns to include (candidate info, scores, metadata) and which question responses to export.