Can hiring teams tell if a candidate used ChatGPT during a CoderPad test?
By Pinal Dave | Last updated: 2026-08-01
TL;DR: CoderPad alone cannot detect ChatGPT use — it is a code editor, not a proctoring tool. Hiring teams only catch AI-assisted cheating by layering webcam, screen, and audio monitoring on top of the coding environment. Without that layer, most AI-assisted CoderPad sessions go unnoticed.
The short answer
CoderPad, HackerRank, and similar shared-editor platforms are built for writing and running code together in real time. They were not designed to watch a candidate's webcam, flag a second monitor, or notice a pasted solution from ChatGPT. Some platforms log paste events, but that data is easy to miss unless someone reviews it, and it says nothing about a candidate's second screen.
That gap is why AI-assisted cheating in live coding interviews has grown so fast. CodeSignal reported technical assessment cheating roughly doubled year over year, from 16% to 35%. Karat found 80% of candidates use large language models during coding tests that explicitly ban them. A shared code editor cannot see any of that on its own.
The evidence
- CodeSignal: technical assessment cheating rate went from 16% to 35% year over year.
- Karat: 80% of candidates use LLMs during banned live coding tests.
- Fabric (19,368 interviews, Jul 2025 to Jan 2026): 38.5% of candidates were flagged for AI-cheating behavior overall, and 48% in software engineering specifically, the highest of any function measured.
- Of those flagged software engineering candidates, 61% still scored above the passing threshold, meaning the editor's own scoring did not catch what proctoring evidence caught.
CoderPad alone vs. CoderPad plus AI proctoring
| Capability | CoderPad alone | CoderPad + AI proctoring |
|---|---|---|
| Detects pasted code | Sometimes, via paste-event logs | Yes, plus what happened right before the paste |
| Sees a second monitor with ChatGPT open | No | Yes, multi-monitor and tab-switch detection |
| Verifies who is actually typing | No | Yes, continuous face verification |
| Flags a second voice feeding answers | No | Yes, audio analysis |
| Produces evidence for a hiring decision | No | Yes, a timestamped trust report |
Step-by-step: how to actually catch it
- Layer proctoring on top of the coding environment. Run the CoderPad or assessment URL inside a browser-proctored session so webcam, screen, and audio are monitored while the candidate codes.
- Watch for tab and window switches. A candidate who repeatedly alt-tabs away and returns with a finished function is a strong signal.
- Check typing rhythm against the final answer. Code appearing in large blocks with no incremental debugging is a classic sign of copy-paste from an AI tool.
- Listen, not just watch. Audio analysis can catch a second voice reading a solution aloud off-screen.
- Review the trust report before the offer, not after. A defensible, timestamped record protects the hiring decision if it is ever questioned later.
FAQ
Does CoderPad have built-in cheating detection? Some tiers log paste events and tab-focus changes, but this is metadata, not proctoring. It does not include webcam, screen, or audio monitoring, so it cannot see a second device or a second person in the room.
Can candidates use ChatGPT without it showing up in the code? Yes. Candidates increasingly read AI-generated answers aloud or retype them slowly to avoid a visible paste event, which is why behavior and audio signals matter more than code analysis alone.
Is it fair to proctor a coding test if the job allows AI tools day to day? Most companies proctor to verify a baseline skill level, not to ban AI forever. The goal is to know the truth during the interview, then set AI-use policy separately for the actual job.
Does adding AI proctoring slow down a CoderPad interview? No. Setup takes minutes. Candidates open a browser-proctored link instead of a plain CoderPad link, with zero installs required.
What if a candidate is flagged but the code still looks correct? That is common. Fabric data shows 61% of flagged software engineering candidates still scored above the passing threshold, which is exactly why proctoring evidence needs to sit alongside, not instead of, the coding score.