Does HackerRank track eye movements?

The worry is that a camera is watching your pupils and quietly scoring you. This section lays out what HackerRank's proctoring options collect and what they do not. The short version is that the feature list is about your browser and your screen, not your gaze.

When an employer turns proctoring on for a test, HackerRank can capture periodic webcam images, log when the test tab loses focus, log full-screen exits, flag copy-paste events, and score your submitted code for similarity against other submissions. Every one of those signals is either an event log or a still image, and each lands in a report a recruiter reads afterwards.

Gaze estimation, pupil tracking, and head-pose scoring are a different class of technology. They need continuous video and a model that emits a per-second attention score. That capability does not appear in HackerRank's proctoring feature list, and the scope of a given test is described in the instructions screen before you start.

Two separate things blur together. Some remote-exam and assessment vendors do sell gaze or head-movement analysis, and a candidate who reads about one vendor tends to assume every platform works the same way. On top of that, HackerRank's webcam snapshots are real, so the camera indicator does light up, and a camera that is on feels like a camera that is analyzing.

The focus log adds to the confusion. If you switch to another tab or leave full screen, the browser reports it and the report gains a timestamped entry. Nothing about that needs a camera, and it catches roughly the behavior gaze tracking would be aimed at, which is likely why the two get conflated.

• Snapshot: a still image a person later looks at.
• Focus event: a browser-level log line, camera or no camera.
• Similarity score: a comparison of your code against other submissions.

A snapshot is a still image, and a person looks at it. If a frame shows you turned toward a second monitor, that is not an eye-tracking score; it is a human forming an impression. The distinction matters, because a human read is contextual and a flag usually invites a question rather than an automatic rejection.

Consider a backend engineer taking a proctored HackerRank assessment for an L5 role at a public cloud vendor. She keeps paper notes beside the laptop and glances down while thinking, so several frames show a downward gaze. Nothing fires automatically. If her submission also scores high on code similarity, though, the reviewer now has two things to ask about instead of one.

The practical conclusion is that a proctored coding assessment is not a place for live assistance. Background on how detection works across platforms sits in the detectability topic hub.

SubcueAI is built for live conversation interviews, not for proctored coding assessments. On macOS and Windows the desktop app captures system audio together with your microphone and shows suggestions in a floating local overlay. The Chrome and Edge extension side panel does live assist from the meeting tab's audio only, so it hears the interviewer and never your microphone. In both cases no bot joins the call and nothing is injected into the meeting page.

The limits are firm. If your screen is shared, recorded, or watched by proctoring software, that situation is outside what the product is designed for, and a company-managed device can enforce rules of its own. A proctored HackerRank test falls squarely in that group, so the useful preparation there is rehearsal rather than assistance. Timed drills against an AI interviewer run on the mock interview page.