AI proctoring for remote team assessment_

Cheating doesn’t show up in a score. Codility layers behavioral, similarity, identity and network signals into one risk level on every candidate report, and your reviewers see the evidence underneath.

87% of candidates rate their experience good or excellent, across 48,000+ post-assessment responses.

homeHiringcontactsvideocamgroup
ScreenSenior backend engineerA. Novak
MH
SummaryDeep diveIntegrityTimeline
A. Novak
REQ-2481 · submitted 14 Aug, 16:42
Total score
94%

Integrity risk

warningModerate
moderate · 2 of 4 categories contributed
content_copy
Behavioral events
Focus change and a paste event on the timeline
warningContributed
difference
Cross-candidate similarity
Checked against other submissions on this task
check_circleNo match
shield
Identity verification
Completed before the assessment started
check_circleVerified
wifi
Network IP check
More than one address seen during the session
warningContributed
2 categories contributed · 6 events on the timeline
Every candidate report carries a risk level and the signals behind it.

Trusted by GitHub, SpaceX, Tesla, EY, LSEG, SAP, Barclays and Citi.

Trusted by GitHub, SpaceX, Tesla, EY, LSEG, SAP, Barclays, Citi.

G2 Leader, Enterprise, Fall 2026
G2 Leader, Mid-Market, Fall 2026
G2 Best Estimated ROI, Enterprise, Fall 2026
G2 Most Implementable, Enterprise, Fall 2026

G2 Fall 2026: Leader in Enterprise and Mid-Market, Best Estimated ROI and Most Implementable in Enterprise.

Paste detection was built for a quieter problem

“I need live proctoring. How do I know if a candidate is using an invisible AI overlay?”

That’s a real review of our own platform, left in 2025. It’s now the most common question buyers bring us.

The gap is evidence. When a score is questioned, your team needs to see how the work was produced.

Paste detection catches a paste. It does not catch a candidate retyping an AI answer line by line from a second screen. Codility offers pattern detection exactly for this problem.

Screen sharing shows you the screen. Tools like Cluely and Interview Coder are excluded from screen capture, so the screen looks clean.

Session recording gives you footage that is difficult to watch at scale. An hour of video per candidate does not scale past a handful of roles.

Exam proctoring platforms were built to monitor a room for universities and certification bodies. They cannot tell you how a piece of code came to exist.

Integrity signals, assembled into something a person can review

Integrity is built into Screen, Interview and Skills Intelligence. Behavioral events, code history, similarity matches, identity checks and network checks feed a single risk level on every candidate report: none, low, moderate or high. Plus pattern detection and invisible AI overlay detection.

Your reviewer opens the level, sees which categories contributed, and drills into the evidence underneath. Codility shows its work. Your reviewer makes the call.

homeHiringcontactsvideocamgroup
ScreenREQ-2481R. Okonjo
MH
SummaryDeep diveIntegrityTimeline
R. Okonjo
REQ-2481 · Python · 60 min · reviewer M. Hallberg
Integrity riskpriority_highHigh

Categories

content_copy
Behavioral
priority_high4 events
difference
Similarity
priority_high1 match
shield
Identity
check_circleVerified
wifi
Network
check_circle1 IP
robot_2
Unauthorized AI tools
priority_high1 signal
Task 2 score
96%
correctness 100 · performance 88

Evidence

session 00:00 to 58:12
04:18Focus left the assessment tab for 3 min 40 sBehavioral
11:02Insertion of 41 lines in a single stepCode history
11:04Paste event, 1,208 charactersBehavioral
37:55Task 2 solution matches submission #C-7719 at 94%Similarity
58:12Session submitted from the same IP as sign-inNetwork

See how the code was written, keystroke by keystroke

Open the Code Evolution Chart and you replay the session as a timeline. Typing bursts, paste events, focus changes and sudden large insertions all sit on the same line, so your reviewer watches the solution take shape.

A 60-minute session becomes reviewable in a couple of minutes.

Your senior engineers stop scrubbing video

homeHiringcontactsvideocamgroup
ScreenR. OkonjoTask 2
MH
SummaryDeep diveIntegrityTimeline

Code evolution chart

task 2 · binary search tree

Typing burstPasteLarge insertionFocus lost
focus lost
+41 lines · 11:02
Paste · 11:04
00:0015:0030:0045:0058:12
Snapshot at 11:02
def insert(self, node, key):+ if node is None: return Node(key)+ if key < node.key: node.left = self.insert(…) return node
Every change, in order, with the proctoring events on the same timeline.
homeHiringcontactsvideocamgroup
ScreenTest mission controlREQ-2481
MH
SortIntegrity riskarrow_downwardScoreSubmitted46 sessions
CandidateScoreRiskCategories
R. Okonjo96priority_highHighBehavioral, similarity
A. Novak94warningModerateBehavioral, network
L. Ferreira88warningModerateBehavioral
S. Iqbal91scheduleLowBehavioral
D. Lindqvist79check_circleNoneNo categories fired
Sort a pipeline by risk level and start where the evidence points.

One risk level instead of a wall of flags

Integrity Risk gives you one risk level across behavioral events, cross-candidate similarity, identity verification status and network checks: none, low, moderate or high. A confirmed similarity match raises it automatically.

  • Deterministic rules produce it, so the same inputs always give the same result.

Your team reviews the sessions that warrant it rather than all of them

Catch the tools that hide from screen sharing

AI overlay apps show candidates answers in a window that screen sharing never picks up. The Codility App, a lightweight desktop app for macOS and Windows, catches those tools on the candidate’s device during the assessment and flags them on the candidate report.

  • Candidates take the assessment inside the Codility App, so the check runs on their device, not just in the browser.
  • The app watches for any program that asks to be hidden from screen capture. Overlay tools can’t stay hidden without that request.
  • So the check still works on tools sold as “undetectable.”
  • It checks the device only. It doesn’t record the screen or log keystrokes.
  • You turn it on per assessment, so candidates only install it where the stakes justify it.
homeHiringcontactsvideocamgroup
ScreenREQ-2481L. Ferreira
MH
SummaryDeep diveIntegrityTimeline
L. Ferreira
REQ-2481 · device check on
warningModerate
desktop_windows
Device check
Codility App · macOS 15.2 · one program requested exclusion from screen capture at 22:41
1 signal
content_copy
Behavioral events
2 focus changes · 1 paste, 340 characters
warning3 events
difference
Cross-candidate similarity
Checked against 4,318 submissions on this task
check_circleNo match
No automatic action taken. Reviewer decision required.
A device signal sits next to the behavioral evidence, on the same report.

See it on a real session

Bring one assessment you already run. We’ll walk the integrity evidence with you end to end.

What AI proctoring can and cannot tell you

No proctoring setup catches everything, ours included.

What a candidate might do What Codility catches in the browser What the Codility App adds (optional)
Pastes an answer in The paste event, where it landed in the timeline, and how large it was Already covered
Retypes an AI answer from a second screen Typing Pattern Detection flags the line-by-line pattern after submission Already covered
Submits work matching another candidate's Similarity Check, which raises the risk level automatically Already covered
Has somebody else sit the assessment ID Verification through a certified third-party provider Already covered
Works from more than one location during a session Network IP check Already covered
Uses a phone off-camera Video and snapshot proctoring, where recording is allowed for the role and region Already covered
Runs a hidden AI assistant on the same machine Nothing, because the tool hides itself from screen sharing Detects the apps running on the candidate's device
Screenshots the task to share it or feed it to AI Can't be prevented in a browser Blocks screenshots during the assessment
Gets help from someone off-camera with recording turned off Not detectable Not detectable

Every signal is evidence, and a reviewer weighs it before anything happens to the candidate.

Two things a reviewer can check for themselves

1. The risk level is deterministic

Fixed rules, no generative model, and the same inputs always produce the same result. Your reviewer can walk an auditor through every step of how a level was reached. A machine-learned score can’t be walked through that way.

2. The device check targets the evasion technique itself

Most platforms watch for known keyboard shortcuts, and the newer overlay tools have already adapted around that. The Codility App looks at how these tools hide from screen capture. Drop that behavior and the tool stops working.

Built to survive a legal review

AICPA SOC for Service Organizations
ISO 27001 Information Security Management certified
GDPR compliant
CCPA compliant

What ships where

Signal Screen Interview Skills Intelligence
Behavioral proctoring Available Available, live to the interviewer Available
Code Evolution Chart Available Available Not available
Similarity Check Available Not available Available
Integrity Risk Available Not available Available
Network IP check Available Not available Not available
ID Verification Premium Premium Not available
Visual and video proctoring Premium Premium Premium
Typing Pattern Detection Available Not available Not available
Cheating apps detection Available Available Not available

What teams say once it’s running

“Codility has streamlined how we evaluate technical roles at Trimble. Beyond a great product and its proctoring features, their customer success team is responsive and dedicated, offering deep-dive support and guidance every step of the way.”

“With Codility, our teams ran 750 candidate tests over 90 days, saving 2,200 hours of interview time. That kind of productivity is like gaining time to launch an entirely new product or enter a new vertical.”

Questions buyers ask

Can a proctor detect AI?

Not reliably. Published research on AI-code detection tools finds critical precision and recall trade-offs and concludes that reliable detection does not currently exist. Codility gives your reviewer the evidence of how the code was produced: typing patterns, the code evolution timeline, paste and focus events, and a device check for hidden overlay tools.

See it on your own pipeline

Bring your questions and one live assessment. We’ll spend 30 minutes in the evidence.

Already a customer? Sign in at login.codility.com

Start a free trial

Bring your questions and one live assessment.