Does Codility detect cheating?
By Aaron Cao · Updated
Yes, mostly after you submit. Codility compares your solution against prior and public code for similarity, logs tab switches and paste events during the test, and can replay how the solution was built. Flags go to a human reviewer, not an automatic rejection.
What does Codility check after you submit?
You want to know whether something is watching you while you work. Mostly it is not; the heavier analysis runs on what you hand in. This section walks the checks in the order they actually happen, so the timing is clear.
During the test, the platform records behavior signals: focus changes when you leave the window, timestamps on large pasted blocks, and the sequence of edits that built each function. Some assessments are configured to require full screen, and some capture a photo for identity.
After you submit, the code goes through similarity comparison against a body of previous submissions and publicly available solutions. That is the check with the most weight, because a well-known problem has thousands of prior answers to compare against.
How does the similarity check actually work?
Similarity comparison is not looking for identical text. It compares structure: the shape of the logic, variable patterns, the order of operations. Renaming variables and reformatting does not move the result much, because the underlying structure is what is being measured.
This is also why the check produces false alarms. A standard algorithm written the standard way looks like every other correct implementation of that algorithm, and there are only so many ways to write a binary search. Reviewers know this, which is why a high similarity score opens a conversation instead of closing a file.
Comparable breakdowns for other assessment vendors are collected on the detectability topic hub.
What happens when a report gets flagged?
The report reaches a recruiter or an engineer with the flag attached, alongside your code, your timings, and the edit history. What happens next depends entirely on the company. Some drop the candidate; many use it as a reason to add a live follow-up round where you walk through your own submission.
A data engineer applying to a logistics company was flagged for a solution that matched a common textbook implementation. The follow-up call took fifteen minutes: they explained the approach, adjusted it live when asked, and the process continued. Being able to defend your own code is what the flag ends up testing.
What SubcueAI records during a live interview, and what it stores afterwards, is set out on the security page.
Where SubcueAI fits, and where it does not
SubcueAI assists live conversations. The native macOS and Windows app captures system audio plus your microphone and shows suggestions in a floating local overlay; the browser extension's Side Panel does live assist for a meeting tab on Chromium browsers, capturing only that tab's audio. No meeting bot joins the call, and nothing is injected into the meeting page.
A Codility assessment is not a conversation, so none of that applies. If the assessment is proctored or your screen is captured, an overlay is inside the capture, and that is a limit rather than a configuration problem. The rounds it is built for are the recruiter screen, the behavioral interviews, and the design discussion that usually follows a passed assessment.
For the assessment itself, the useful preparation is timed practice, which runs on the mock interview page.