Can HackerRank Detect ChatGPT, Copilot, or Cluely?

It depends on which HackerRank mode the employer enabled, and on which of two different questions you are asking. Detecting the tool itself is something only Proctor Mode and Desktop App Mode attempt, by analyzing screenshots or blocking applications. Detecting the tool's output is the job of HackerRank's AI plagiarism model, which an employer can enable on any test and which HackerRank rates at 85% precision. As of September 9, 2026, HackerRank's documentation names Cluely and "LLM applications" as things Desktop App Mode blocks, and lists "External AI coding assistants" and "invisible overlay applications" as screenshot-analysis targets.

No HackerRank mode reads what ChatGPT told you. Every mode above the default can collect evidence that something did.

Two detections, not one

Detecting the tool on screen or on the machine

HackerRank's Proctor Mode documentation says the system captures the entire screen every 15 seconds, dropping to five seconds after suspicious activity, and analyzes those screenshots for "External AI coding assistants, such as third-party AI tools or copilots, not integrated within HackerRank's IDE" and for "Specialized browser extensions or invisible overlay applications that assist in answering coding questions". Each screenshot is classified as suspicious or not and shown to the recruiter with an explanation.

Desktop App Mode does not wait for a screenshot. HackerRank's Desktop App Mode documentation states that the app "Blocks the use of other applications during the test, including tools such as Cluely, video conferencing, messaging applications, or LLM applications", and that it prevents remote access and screen sharing.

Secure Mode and the default test do neither. They record window exits, full-screen exits, and paste events, and nothing else about what is running on your computer.

Detecting the output in the code

HackerRank's AI Plagiarism Detection documentation describes a model that scores "Code writing patterns", "Time taken to solve the problem", "Copy-paste activity", and "Tab-switching behavior", with two named patterns: "External Copy-Paste" and "Suspicious Code Resetting", meaning large deletions followed by irregular typing. HackerRank reports "an overall precision of 85%, meaning flagged sessions are correct 85% of the time." Flags are rated High or Medium, and a reviewer can replay the typing in the Keystroke Codeplayer.

This layer does not care which tool produced the code. It cares how the code arrived in the editor. HackerRank also notes the model is weak on "questions requiring minimal effort to achieve a full score or when solutions consist of very few lines of code."

Tool by tool

ChatGPT in another tab, or on a phone

Switching to a ChatGPT tab is logged on every HackerRank test, because tab proctoring is on by default and reports "Out of Window Duration" and "Number of Window Exits". Secure Mode adds an immediate warning. In Proctor Mode, the ChatGPT window is on the screen being captured, so it lands in a screenshot and in the AI-assistant category.

A phone is outside the browser's view. HackerRank's own article on tab proctoring lists "Smartphones, tablets, or additional computers used for research" among the things that layer cannot see (Tab-Proctoring: What It Catches and What It Misses). Two things still apply: Desktop App Mode flags "mobile phones and tablets in the candidate's webcam feed", and anything retyped from the phone still passes through the plagiarism model.

GitHub Copilot in a local editor

Copilot never runs inside HackerRank's editor, so its output has to be moved in. Pasting is an "External Copy-Paste" event, logged with its content on every test and blocked outright in Secure Mode and Desktop App Mode. Retyping avoids the paste event and produces the timing and code-arrival pattern the model scores instead.

Whether Copilot is cheating is a rules question. HackerRank's AI add-on documentation lets employers provide an AI assistant inside the IDE with reviewable prompts. If the test says no external tools, Copilot is an external tool.

Cluely

Cluely is the one product HackerRank names. Desktop App Mode blocks it as an application. In Proctor Mode, an overlay on the captured screen falls under the "invisible overlay applications" category of screenshot analysis. Whether a given overlay appears in a given screenshot depends on how the capture is taken, which HackerRank does not publish, and the plagiarism model runs on the typed result either way.

Interview Coder, ParakeetAI, and other overlays

HackerRank names the category, not each product, and Desktop App Mode blocks other applications generically. Nothing in HackerRank's documentation supports calling any assistant undetectable, and Control, a desktop AI interview assistant for interviews, assessments, and screen-share workflows, makes no HackerRank compatibility claim. The reasons a screen-share test proves less than vendors imply are laid out in why invisible AI tools are not a HackerRank bypass.

When AI is allowed on HackerRank

Some HackerRank tests and interviews permit AI, through the employer-enabled assistant inside the platform. In those sessions the prompt transcript is part of what gets reviewed, as described in what HackerRank keeps from your AI prompts. Permission comes from the rules on the consent screen, never from a tool's marketing, and when AI is allowed in a technical interview covers how to confirm it in writing.

What this means for candidates

Identify the mode first. The full list of what each mode records is in what HackerRank proctoring detects and records. If the test prohibits external AI, use AI to prepare beforehand and take the test without it; the tooling question does not change the rules question. If AI is permitted, use the tool the employer provided, keep the code something you can explain line by line, and expect the prompts and the typing replay to be read by a person.