Can HackerRank tell if you use ChatGPT?

The worry behind this question is usually specific: you have ChatGPT open in a second tab and you want to know whether the test window can tell. This section lists what HackerRank records. The short version is a handful of browser-level signals plus an optional webcam, and none of them reads the contents of another tab.

What a HackerRank assessment logs:

• Focus and tab changes, meaning the moment you left the test window and the moment you came back
• Paste events into the code editor, including how much text arrived at once
• Full-screen exits, when the assessment was configured to require full screen
• Webcam snapshots, only when the company turned on proctoring

A browser tab cannot read the tabs beside it. HackerRank knows that focus left the assessment; it does not know whether you opened ChatGPT, a documentation page, or your email. Platform-by-platform breakdowns live in the detectability cluster.

The risk is not in opening ChatGPT. It is in what you do with the answer.

HackerRank runs code-similarity scoring on submissions, comparing your solution against other candidates' submissions and known public sources. Language models are consistent by design: ask several people to have ChatGPT solve the same standard array problem and the submissions converge on the same structure, the same variable names, and the same comment style. That clustering is precisely the pattern similarity scoring exists to surface.

Pasting compounds it. A single paste that drops a complete solution into an empty editor is a stronger signal than the code itself, because it shows the answer arrived whole instead of developing.

Note what none of this establishes. A similarity flag says two submissions look alike. It does not say ChatGPT wrote either one. Tools that claim to identify AI-written text are unreliable in both directions, and a careful reviewer treats them as a reason to ask questions rather than as evidence.

A second device sits outside the system entirely. If you read ChatGPT on your phone, there is no browser signal to log, because nothing happened in the browser running the test.

That gap is narrower than it sounds once proctoring is on. Webcam snapshots do not detect a phone, but they show a person looking down and away at intervals, and a human reviewer draws the obvious conclusion. On a company-managed laptop, the employer's own device software may see considerably more than HackerRank does.

Consider a backend engineer taking a timed HackerRank screen for a payments team. They solve the first two problems in the editor and work the third through ChatGPT on their phone. The HackerRank log shows a clean session. The webcam snapshots show attention off-screen for most of the final stretch, and the third solution reads nothing like the first two. Nothing was detected, and the report still looks odd. Rehearsing the rounds that follow with mock interview practice is the part that actually compounds.

SubcueAI is built for live conversational interviews, the ones where a person asks you questions over Zoom, Google Meet, or Microsoft Teams. The desktop app on macOS and Windows captures system audio and your microphone and shows suggestions on a local overlay; the browser extension Side Panel does the same for a call running in a Chromium tab, capturing that tab's audio only, which means the interviewer and not you. No bot joins the call and nothing is injected into the meeting page.

A HackerRank coding assessment is a different situation. Nobody is speaking, so a live assistant has nothing to listen to, and bringing outside help into an assessment breaks the terms you accepted when you started it. We do not claim any tool is undetectable there. Screen sharing, screen recording, proctored environments, and company-managed devices all sit outside what a local overlay controls.

The preparation that works for a coding screen is the unglamorous kind: practice the problem types until you can write them unaided. The capture and latency details are covered in the how it works cluster.