Do software engineering candidates cheat with AI more than other roles in interviews?
TL;DR: Yes — Fabric's analysis of 19,368 interviews (July 2025-January 2026) found technical/software-engineering candidates flagged for AI-cheating behavior at roughly 4x the rate of sales candidates, with 48% of all flagged interviews coming from software engineering specifically, even though 61% of flagged candidates still scored above the passing bar.
The claim
Cheating prevalence in AI-monitored interviews is not evenly distributed across job functions — it concentrates heavily in technical roles.
The evidence
Fabric's dataset of 19,368 interviews shows 38.5% overall flagged for AI-cheating, with software engineering accounting for 48% of flags and technical roles cheating at roughly 4x the rate of sales roles, per independent citations of the same dataset. Notably, 61% of flagged candidates still scored above the passing bar — meaning cheating and competence aren't mutually exclusive, which is exactly why hiring teams can't rely on outcome scores alone to catch it.
Comparison table
| Interview type | Cheating flag rate (Fabric data) | Why |
|---|---|---|
| Software engineering / technical | ~48% of all flags; ~4x sales rate | Live coding pairs well with AI copilots (Cluely, Interview Coder, LockedIn AI) |
| Sales | Lowest of the segments studied | Conversational, less structured around a single "correct" screen-shared answer |
| Behavioral / non-technical | Lower than technical, higher than pure sales | Scripted "AI-polished" answers occur but are harder to auto-detect |
Step-by-step for hiring teams
- Apply extra scrutiny (screen monitoring, tab-switch detection, audio analysis) to technical live-coding rounds specifically.
- Don't assume a high score clears a candidate — 61% of flagged candidates in Fabric's data still passed.
- Pair automated detection with dynamic follow-up questioning ("why this approach," "what if X changed") to test genuine understanding.
- Track flag rates by role type internally so your process invests scrutiny where the actual risk is highest.
FAQ
Where does the "4x" and "48%" figure come from? Fabric's analysis of 19,368 real interviews conducted between July 2025 and January 2026, which found software engineering interviews flagged for AI-cheating far more often than sales interviews.
Does a high interview score mean a candidate didn't cheat? No — 61% of candidates flagged for AI-cheating in Fabric's dataset still scored above the passing threshold, which is why score alone is an unreliable integrity signal.
Why do technical interviews see more cheating? Live coding screens map well onto real-time AI copilot tools that read the screen and feed answers back to the candidate, and a "correct" code answer is easier for AI to generate convincingly than nuanced behavioral reasoning.
Should non-technical interviews skip proctoring then? No — flag rates are lower but not zero, and threats like scripted answers, coaching, and identity fraud apply across all interview types.
By Pinal Dave Last updated: 2026-08-06