Persona angle
AI detection for recruiters
A detector can help you decide where to spend attention. It cannot decide who gets the job. This page is the workflow version of that sentence.
Verified 2026-08-05Use scores for triage, not rejection
Route high-score applications to a slower human read. Do not wire the percentage into an automatic reject rule. The cost of a false positive is a missed hire; the cost of a false negative is usually ten more minutes of reading.
Pair detection with substance checks: can the candidate talk about the projects on the CV? Do work-sample answers match the application voice?
Bias you cannot ignore
Detectors over-flag non-native English. If your pipeline is global, a blunt score threshold becomes a language test dressed up as an integrity test. That is both unfair and easy to challenge.
Train reviewers to open sentence-level evidence and to treat "Most likely AI" and "Too close to call" as different states. The middle of the scale exists on purpose.
A process that survives scrutiny
Write down that scores are indicators. Keep the extract and the report if you act on them. Offer candidates a chance to discuss flagged passages. Never lead with "our AI proved you cheated".
For CVs and cover letters specifically, read the dedicated angles: those formats need looser priors than a long essay.
Questions
- What AI score should reject an application?
- None as a sole rule. Use bands to prioritise review. Decide on substance, interview signal and work samples.
- Can a candidate challenge an AI-detection rejection?
- They can challenge the fairness of the process. A score without human review is a weak position. Document how you used the tool.
