Cover letters written by AI: what to do about it in hiring
Anyone can now have a cover letter generated in a minute, so its value as a signal has dropped. Detection shows you where not to spend time, but the fix is to change the task — to ask about things that cannot be written without knowing the candidate.
Why the filter stopped working
A letter that used to show interest and an ability to express oneself now mostly shows that the candidate can write a prompt.
Punishing that makes no sense. Adapting the process to it does.
Where detection helps in hiring
In larger hiring rounds, as triage. When a letter is generated and empty at the same time, you save yourself a round that would lead nowhere.
The score alone must not be a reason for rejection, though. The candidate may have used AI for language editing, which is legitimate.
What works better
A short task tied to your company. A question whose answer cannot be generated without knowing the specific context.
Fifteen minutes of conversation says more than three pages of a letter, whoever wrote it.
FAQ
Can we reject a candidate because of a detection?
It is not defensible. Detectors get it wrong, and the decision would rest on unreliable grounds.
Should we tell candidates that we check?
Yes. A transparent rule is fair and at the same time reduces the number of plainly machine-made applications.
How does this sit with GDPR?
You are checking a document the candidate sent you. As with other material, you handle it according to your own processing policy.
Try it on your own text
Paste a text or upload a file and look at the score and at the specific sentences that came out suspicious.
Run a detection