Reviewable evidence. Human accountability_
When every candidate has access to AI tools and screen-capture-invisible cheating apps, a test score alone is not enough. Codility delivers layered trust controls that give reviewers evidence they can act on.
Assessment integrity is harder
than it used to be
Candidates now have access to AI coding assistants, screen-capture-invisible tools, and solution databases. Whether you enable AI assistance or switch it off, a single integrity check is no longer sufficient. Your process needs layered, reviewable evidence at every stage.
AI tools rewrite candidate behaviour
Candidates can generate, paste, or retype AI-generated solutions. Traditional paste detection alone does not surface these patterns.
Cheating apps evade screen sharing
Apps like InterviewCoder and similar overlay tools request exclusion from screen capture, so they stay invisible to standard proctoring.
Manual review does not scale
Reviewing video for every candidate is costly and slow. Teams need aggregated signals that surface the sessions worth reviewing.
One signal is not enough
No single check covers every scenario. A candidate might pass paste detection but fail identity verification. Only layered evidence gives the full picture.
From a number to reviewable evidence
| Without layered integrity | With Codility Proctoring & Integrity |
|---|---|
| A result with no supporting evidence | Behavioural, identity, and similarity signals in one reviewable record |
| Paste detection only, blind to AI retyping patterns | Typing Pattern Detection flags line-by-line retyping from external AI sources |
| No identity verification before the assessment starts | Government-issued ID verification before the first task |
| Manual video review for every candidate | Deterministic Integrity Risk surfaces the sessions that need review |
| No visibility into screen-capture-invisible apps | Cheating apps detection surfaces apps that exclude themselves from screen capture, for human review |
| No way to verify the candidate understood what they submitted | AI Follow-Up Questions ask candidates to explain their reasoning, on the spot |
| A downloaded report that leaves out AI and integrity context | One PDF report with integrity risk, proctoring summaries, and AI activity together |
Layered integrity signals across every dimension
Behavioural Proctoring
Automatic detection of suspicious signals: code pasted from external sources, task description copied, time away from the IDE, and abnormally fast completion. In Interview, signals surface live to the interviewer during the session.
Code Evolution Chart
Visual timeline of how the candidate’s code changed over time. Proctoring events are overlaid on the code history, surfacing sudden large insertions and other suspicious patterns.
Similarity Check
Automated cross-candidate plagiarism detection. Compares submissions against other candidates’ solutions and flags matches for manual review. Positive matches automatically elevate Integrity Risk to High.
ID Verification
Candidates verify their identity with a government-issued ID before the assessment begins. Results feed into Integrity Risk under the Identity and Network category.
Data handling reference →Visual & Video Proctoring
Periodic webcam and screen snapshots during Screen assessments. Full webcam, screen, and audio recording is also available. Interview includes session recording and full transcript as a premium feature.
Data handling reference →Network IP Check
Checks whether a candidate used multiple IP addresses during an assessment. A single IP means no risk. Multiple IPs flag medium risk, adding another data point to the Integrity Risk calculation.
Typing Pattern Detection
Identifies candidates who retype AI-generated solutions line by line from an external source. Analyses typing behaviour after submission, with zero disruption during the assessment. Fully passive, requires no setup, and covers most coding task types.
AI Follow-Up Questions
Automatically generated questions that ask candidates to explain the choices and reasoning behind their submitted code. Surfaces candidates who can produce AI-generated output but cannot defend it. Intentionally not scored, to keep the signal human-judged and free of bias.
Cheating apps detection
via the Codility Desktop App
A lightweight desktop app candidates install before the assessment. It is designed to surface apps that request exclusion from screen capture, the technique tools like InterviewCoder and similar overlay tools use to stay invisible to standard proctoring, and hand that evidence to a human reviewer. Codility does not auto-fail a candidate on a detection. It works alongside browser and video proctoring, and results appear in the report integrity widget with live notifications for interviewers.
Layered signals. One reviewable risk level.
Integrity Risk aggregates behavioural, identity, network, and similarity signals into a single qualitative risk level: None, Low, Moderate, or High. Scoring is deterministic, not generative or machine-learning based. It appears on every candidate report and in the Test Mission Control view.
This is decision support, not a verdict. Reviewers see which signal categories contributed to the risk level and can drill into the underlying evidence. The final call always stays with a human. AI Follow-Up Questions feed into this evidence but are intentionally left out of the scoring itself, to avoid bias.
Configure once. Evidence collects automatically.
Configure integrity settings
Choose which proctoring levels to enable when creating the test: behavioural signals, ID verification, visual snapshots, or full video recording. Settings lock once the first candidate is invited.
Candidates complete the assessment
Integrity signals are captured automatically during the session. Behavioural events, code changes, snapshots, and video are recorded without disrupting the candidate experience.
Review the evidence
The Integrity Risk widget on each candidate report shows the risk level and the underlying signals. Drill into code evolution, video playback, or download the full PDF report for a complete, offline record.
Where each signal is available
| Feature | Screen | Interview | Skills Intelligence |
|---|---|---|---|
| Behavioural Proctoring | ✓ | ✓ | ✓ |
| Code Evolution Chart | ✓ | ✓ | |
| Similarity Check | ✓ | ✓ | |
| Integrity Risk | ✓ | ✓ | |
| ID Verification | Premium | Premium | |
| Visual & Video Proctoring | Premium | Premium | Premium |
| Network IP Check | ✓ | ||
| Typing Pattern Detection | Preview | ||
| AI Follow-Up Questions | Preview | Preview | Preview |
| Cheating apps detection | Preview | Preview | |
| Redesigned PDF report | Preview | Preview |
Interview recording and transcript available as a premium feature. Behavioural proctoring signal coverage may vary by Interview environment type.
Evidence you can act on after every assessment
Every completed assessment produces structured integrity data. Reviewers see the full picture without watching every recording, and can take the full picture with them in one download.
Integrity Risk widget
Deterministic risk level on every candidate report. See which signal categories contributed and drill into the evidence.
Code evolution timeline
Visual history of every code change, overlaid with paste events, focus changes, and other behavioural signals.
Similarity report
Cross-candidate plagiarism status with simplified labels. Positive matches automatically elevate Integrity Risk to High.
ID verification status
Verified, failed, or not attempted. Status feeds directly into the Identity and Network signal category.
Video and snapshot playback
Webcam, screen snapshots, and full session recording viewable alongside the code timeline.
Typing pattern flag
Risk flag when code was entered in a line-by-line pattern that suggests retyping from an external AI source. Passive, post-submission.
AI follow-up responses
Candidate answers to AI-generated reasoning questions, attached to the submission. Reviewers judge depth of understanding directly. Not scored, by design.
Detected apps report
Apps that requested exclusion from screen capture, with detection times, surfaced in the candidate report for human review.
Redesigned PDF report
Thoroughly redesigned in June 2026. One faithful, complete download with Integrity Risk, the integrity widget, full proctoring summaries, code health, AI Assistant and AI Copilot transcripts, and skills evaluations together.
codility.com