How to Detect AI Cheating in a UX Designer Teams Technical Screen
UX designer Teams technical screens typically involve a live design critique or a whiteboard usability-heuristics discussion. AI cheating shows up as a candidate reciting LLM-generated heuristic evaluations off a second screen — fluent, textbook-correct, but generic and unable to connect to the specific screen or flow you show them. Neuroxa's AI Meeting Proctor tracks gaze and screen-share activity through the call.
| Threat Model | Observable Tell | Confidence |
|---|---|---|
| LLM chat window open generating heuristic-evaluation answers | Fixed-point gaze before a fluent, generic answer | High |
| Screen-share briefly reveals a second application | Application-switch event logged in Teams screen-share metadata | High |
| Memorized "Nielsen's 10 heuristics" recitation regardless of the specific screen shown | Answer doesn't reference details specific to the design you presented | Medium-High |
| Audio delay suggesting a remote collaborator whispering answers | 2-4 second consistent delay before answers | Medium |
Interviewer script: "I'll share my screen with a real flow from our product — critique it live, and I'll push on specific choices." Point to one specific element (e.g., "why do you think this button is here?") and see whether the answer is grounded in what's on screen or generic.
Evidence to capture:
- Full Teams call recording with gaze overlay
- Application/window-switch events during screen share
- Response specificity against the actual screen/flow shown
- Response latency per question
- Flagged-moment screenshots for reviewer sign-off
Neuroxa product: AI Meeting Proctor — live Teams call monitoring with gaze tracking and screen-share application-switch detection.
FAQs
Isn't knowing heuristic frameworks by name a good sign? Naming a framework is fine; the flag is a generic recitation that doesn't connect to the actual design shown on screen.
What if the candidate sketches on paper during the critique? Paper sketching without a fixed screen-gaze pattern is a different, non-flagged behavior from reading off a second monitor.
Should we ask candidates to critique something unfamiliar to reduce memorized answers? Yes — using your own product's real screens (rather than a generic sample app) makes memorized, generic answers stand out more clearly.
How should a flag be handled? Review the recording and ask a live follow-up question in a later round if you're unsure — never reject on the flag alone.
Related: UX Designer Zoom Panel Interview · UX Designer Contractor/Offshore Onboarding Verification · Paralegal Teams Technical Screen · Cloud Architect Teams Technical Screen
Secure your design critique screens with Neuroxa.ai AI Meeting Proctor.