What this catches, in practice.

A senior backend candidate breezed through a first-round technical screen — clean, structured answers to system-design questions that usually trip people up. Candit flagged two of them mid-call at 91% and 94% confidence, each tied to a sudden jump from halting to polished phrasing, matched against a live build of a popular stealth copilot. One live follow-up question, asked with no notice, and the candidate's answer fell apart. The role went to the second-place candidate instead.

Running high volume for an enterprise client, one team noticed a string of candidates giving unusually polished answers to open-ended behavioral questions. Turning on per-candidate authenticity scoring surfaced the scale of it — roughly a third of interviews that month included at least one instance above the 40% threshold, concentrated in behavioral rather than technical questions. That let the team retrain interviewers to ask tighter follow-ups exactly where the pattern showed up most.

A small executive search firm added Candit mainly for the meeting notes and ATS sync — AI Detectability ran quietly in the background for months before it ever flagged anything. When it did, on a final-round finance director candidate at 88% confidence across two separate answers, the team already trusted the evidence enough to act on it immediately. They haven't paid out a bad-hire refund since.