Can a candidate use an agentic AI browser like Perplexity Comet to cheat during a live interview?

TL;DR: Yes, in principle — agentic AI browsers like Perplexity Comet are built to read and act on whatever is on screen, which makes them usable as a live cheating aid the same way overlay tools like Cluely are. Perplexity's own CEO has publicly reacted to reports of Comet being used to cheat, confirming the misuse case is real, not hypothetical.

By Pinal Dave | Last updated: 2026-08-05

The claim

Agentic browsers are a newer category than dedicated cheating overlays, but the underlying risk is the same: any tool that can read a question and generate an answer in real time, on the same device the candidate is using for the interview, can be pointed at an interview instead of a homework assignment.

The evidence

Perplexity's CEO was reported reacting to a case of the company's free AI browser being used by a student to cheat, publicly saying "absolutely don't do this" — confirming the tool has already seen real-world misuse for cheating in assessed contexts. The mechanism transfers directly to interviews: an agentic browser open in a second tab or window can read an interview question posed on screen and surface an answer, functionally identical to how Cluely-style overlays work.

Comparison: agentic AI browsers vs. dedicated cheating overlays

Factor	Agentic AI browser (e.g., Comet)	Dedicated overlay tool (e.g., Cluely)
Purpose-built for cheating	No — general productivity tool	Yes, often marketed for exactly this
Requires visible tab/window switch	Yes, typically	Often designed to avoid detection in screen shares
Detectable via tab-switch monitoring	Yes	Depends on tool design
Detectable via gaze/behavior monitoring	Yes	Yes

Step-by-step: reducing the risk

Use browser lockdown for any assessment delivered inside a proctored browser session, closing the tab-switching gap entirely.
For live meeting interviews, keep AI Meeting Proctor active to catch gaze deviation and answer-timing patterns regardless of which tool the candidate used.
Enable multi-monitor and tab-switch detection so a second window running an agentic browser is flagged even without content-level access to it.
Don't rely on banning specific tools by name — the detection layer should catch the behavior, not the brand.

FAQ

Is Perplexity Comet more dangerous than a standard cheating overlay? Not inherently — it's a general-purpose tool being repurposed, similar in effect to any AI copilot a candidate could open in a browser tab.

Can tab-switch detection alone stop this? It catches the switch itself; combining it with gaze tracking and behavior analysis catches subtler cases where the candidate doesn't fully switch away from the interview window.

Does this apply to other agentic browsers too? Yes — any AI browser with read-and-respond capability carries the same risk profile, regardless of vendor.