Screen Overlays in Interviews: What Hiring Managers Need to Know

By Vaibhav Devere, Founder, Zero Assist · 2025-04-14 · 5 min read

The New Normal

A 2024 survey by HackerRank found that over 70% of developers admitted to using AI tools during take-home assessments. The number for live interviews is lower — but growing fast as overlay tools make it harder to detect.

ChatGPT and Claude are not inherently cheating tools. But when a candidate alt-tabs to paste your question and alt-tabs back with a polished answer, they have changed the nature of the evaluation.

How to Spot It Without Technology

There are behavioral signals worth watching for during live technical interviews:

  • Unnaturally long pauses before rapid-fire correct answers. Humans who understand a problem narrate their thinking incrementally. A 45-second pause followed by a complete, correctly-structured solution is unusual.
  • Generic explanation language. ChatGPT answers tend to use consistent phrasing — "certainly," "let me break this down," "firstly." Candidates parroting AI output often use the same register.
  • Perfect code on the first attempt. Real engineers make typos, reconsider variable names, and iterate. A candidate who produces working code without any visible struggle warrants follow-up questions.

Follow-Up Questions Are Your Best Tool

No AI tool can prepare a candidate for genuinely spontaneous follow-up. If a candidate's solution looks too clean, pivot:

  • "Walk me through how you would test this."
  • "What happens to your approach if the input is 10x larger?"
  • "Tell me about a time you encountered something similar in production."

A candidate who used AI to generate the answer will struggle here. Someone who genuinely understands the solution will not.

When You Need Technology

Behavioral signals are useful but imperfect. A well-prepared candidate who has practiced with AI outputs can mimic natural delivery convincingly. That is where process-level monitoring becomes valuable — it removes the ambiguity by telling you whether a cheating tool was actually running, not just whether the candidate seemed suspicious.

Combining both layers — interviewer intuition and real-time monitoring — is the most reliable approach available today.