Document angle
Detect AI writing in a CV — carefully
Résumés are short, formulaic and often written in a second language. That is exactly the shape that raises AI scores even when a person wrote every line. Use the detector as a reason to read closer, never as a reason to reject.
Verified 2026-08-05Why CVs flag so easily
A CV rewards the same patterns a detector reads as machine-like: parallel bullet structure, stock verbs (led, drove, owned), narrow punctuation, and little personal voice. Recruiters asked for that shape for decades before ChatGPT existed.
Non-native English writers are hit harder. Published work on commercial detectors has repeatedly shown elevated false-positive rates on second-language prose. A global hiring pipeline that auto-rejects on score will quietly discard qualified people.
How to check a CV usefully
Extract the text first if you only have a PDF, then run the AI detector. Open the sentence-level view. Ask which bullets carry the signal: generic summary paragraphs flag more often than specific metrics and named tools.
If the score is high, read the document. Look for concrete projects, employer-specific detail and inconsistent voice across roles. A generated CV is usually bland in the same way everywhere; a human CV is uneven.
What not to do with the score
Do not set a percentage cutoff in your ATS. Do not tell a candidate the tool "proved" they used ChatGPT. A detection score is an indicator, not authorship evidence, and it will not stand up as the sole basis for a hiring decision.
If you need process language for your team, use the recruiter angle: it is written for screening workflows rather than for candidates checking their own file.
Questions
- Can I reject a CV because an AI detector flagged it?
- You can, but you should not treat the score as proof. CVs are structurally prone to false positives. Use the score to decide where to read harder, then judge the substance.
- How do I tell AI-generated CV text from a template?
- You often cannot from a score alone. Templates and generators both flatten rhythm. Specificity, verifiable claims and uneven section quality are better tells than the percentage.
- Can I check a CV PDF?
- Yes. Extract text from the PDF first, then run the detector on the extract. Scanned CVs need OCR.
