Checking portfolios and assignments for AI in hiring

Checking for AI use in hiring has become common. What matters is not declaring that AI was used but finding out what made it look so, and treating the candidate fairly.

Decide first what you are asking

The same result means something different for an assignment that banned AI tools and one that asked for disclosure. Stating the check criteria in the job posting removes much of the later dispute.

Image portfolios

Design and photography portfolios are read with the image detectors and provenance signatures. A generating tool's record or signature is strong evidence; without one, read the detector result as a range. Work partly retouched with AI can be missed.

Writing and cover letters

Text detection is in beta and flags a fair number of human texts, especially translated and short ones. Do not reject a candidate on a text result alone.

Using it fairly

If a result raises doubts, ask the candidate to explain the process in the interview or to share drafts and original files. Inconclusive means no firm call can be made, so it should never count against anyone.

Frequently asked

Can I reject a candidate if AI is detected?
We advise against it. Results are probabilistic and include false alarms, so decide together with other checks.
Can I show the result to the candidate?
You can hand over the report with a share link. With the grounds open, explanation and appeal are easier.
Can I check many applications at once?
Upload a ZIP and each file inside is scanned, with the results shown together.
Are candidates' files kept?
Files are deleted after analysis; only the hash and report remain.

Read next

This is general guidance, not legal advice. A scan result is a probabilistic assessment, not proof.

Try a file