How Goodfit measures candidate integrity

How Goodfit Flags Possible Integrity Issues

When you review a candidate's assessment or interview, you may see one or more "integrity flags." Here's what they mean and how to use them.

A flag is not proof of cheating. It just means something happened that's worth a quick human look. Always check the details before making a decision.

The types of checks we run

1. AI / Script-Reading Detection (Voice Interviews)

We analyze the interview transcript for signs the candidate may be reading a script or an AI-generated answer — things like unnatural pauses, oddly "listy" jargon-heavy answers, or answers that don't match the question asked.

Important: we do not flag someone just for speaking confident, fluent, or well-structured English — including natural Indian English/ESL patterns. Only clear transcript-based evidence triggers a flag.

Each interview gets a text integrity score. Higher scores mean stronger evidence of an issue; a case gets sent for human review once it crosses a set threshold. A low score doesn't guarantee a clean interview — it just means we didn't find strong enough evidence to flag it.

2. Proctoring Events (Assessments)

During browser-based assessments we watch for behaviors like switching tabs, exiting full screen, copy/paste attempts, using dev tools, the camera losing the candidate's face, or extra people/voices being picked up.

These are grouped roughly as:

• High concern: tab switching, exiting full screen, dev tools open, extra people or voices detected
• Medium concern: copy/paste/print attempts, looking away for a while
• Low concern: blocked shortcuts, right-click attempts, a brief moment where the face wasn't visible

If enough of these events pile up in one attempt, it gets marked as proctoring flagged so you know to take a closer look.

3. Voice-Video Match (select interviews)

For some voice interviews, we also check whether lip movement lines up with the audio, as an extra signal that the person on camera is the one actually speaking.

How to Read a Flag

Don't judge a candidate off one single flag. The strongest cases for concern usually combine multiple signals — for example, several proctoring events happening together, or a script-reading flag alongside unusual proctoring activity. One isolated low-level event (like a single right-click) usually isn't meaningful on its own.

Recommended steps when reviewing a flagged candidate:

1. Identify what type of flag it is (behavior / answer-pattern)
2. Look at the timestamped details and any flagged transcript excerpts
3. Check if there's more than one supporting signal
4. If unsure, consider re-running the interview or assessment
5. Use your own judgement for the final hiring call — Goodfit surfaces signals, it doesn't make the decision

Need Help Interpreting a Flag?

Reach out to Goodfit support if:

• You want help understanding an AI-answer flag
• Proctoring events seem unusually high (could be a technical glitch)
• You want your integrity settings tuned for a specific role

When you contact us, please share: candidate name, job/assessment name, attempt date & time, and which signal you'd like reviewed (plus a screenshot if you have one).