What a detector can and cannot show in hiring
A detector reads only the finished text. It estimates how closely the wording resembles AI writing and, in Textumi, marks the passages that weighed most in that judgment.
It cannot tell you whether the candidate wrote the letter, used AI for a first draft, asked a friend for help or ran the text through a grammar tool. It knows nothing about the person's skills, experience or motivation.
Every result is an estimate from the style of the text, not proof of authorship. Used carefully, it can start a question; used as a filter, it would reject people for the way their writing happens to read.
Why cover letters are flagged so often
Cover letters are one of the hardest kinds of text for any detector. They follow a template, use the same polite formulas and are polished many times, and predictable phrasing is exactly what detectors associate with AI.
- Applicants are taught standard openings and closings, and career guides recommend the same phrases to everyone.
- Writing in a second language often produces safer, more regular sentences. Stanford researchers reported in 2023 that detectors flagged essays by non-native English writers more often.
- Short texts give a detector little to work with. Under 50 words Textumi shows “Too little text to judge” instead of a label.
- Grammar tools and careful editing smooth out the small irregularities that make writing look personal.
So a flag on a cover letter carries even less weight than a flag on an essay. Treating it as evidence would hit careful writers and international candidates hardest.
A fair way to use a detector in recruitment
If you use a detector at all, use it to prepare a conversation, never to sort candidates.
- Decide and write down your position on AI help before you open applications. Many employers accept AI-assisted letters as long as the content is true; say so in the job ad if that is your view.
- Never reject, rank or shortlist a candidate on a detector result. A person reads every application and decides on the substance.
- If a letter reads as generic, ask about it in the interview: what drew the candidate to the role, what a project on their CV involved, why they chose this team.
- Judge skills with work that shows them directly, such as a short task, a portfolio or a conversation with the team.
- Tell candidates if you check application texts with any automated tool, and keep the results out of their file unless your process requires otherwise.
A cover letter written with AI says little about whether someone can do the job. A specific, honest answer in an interview says a lot.
Rules to know in the EU
AI systems intended to be used for recruiting or selecting people, including for the analysis and filtering of job applications and for the assessment of candidates, are referred to as high-risk AI systems under Annex III of the EU AI Act (Regulation (EU) 2024/1689). Organizations that provide and use such AI systems are subject to obligations, including human oversight.
The GDPR, Regulation (EU) 2016/679, gives people the right in Article 22 not to be subject to a decision based solely on automated processing that produces legal effects for them or similarly significantly affects them, with limited exceptions.
What these rules require of your process depends on how a tool is used and on national law. This page is not legal advice: if a tool filters or scores applications, ask your data protection officer or a lawyer.
If you are applying for a job
Using AI to draft a cover letter is common and often accepted. What matters is that every claim in it is true and that you can talk about it in an interview.
Make the letter specific: name the role, the company and one or two concrete things you did that fit it. That makes it better for the reader, whatever any detector says.
If an employer asks you not to use AI, respect that, and keep your drafts in case you are asked how you wrote the text.
Make AI text sound naturalHow to tell if a text was written by AI
Checking an application text in Textumi
Paste the letter into the workspace or import it from a DOCX or PDF file. Textumi shows one of five labels, from “Human-written” to “AI-generated”, and with an account it marks the passages that weighed most.
Submitted texts and results are stored encrypted. Detection runs on Textumi's own models and computing capacity. On Individual and Professional, checks stay in your private History for 30 days unless you delete them earlier. On the other plans you can open a result for about an hour after it finishes; after that it is no longer available to you, and it is deleted within 7 days. Until then it is kept encrypted, and only Textumi's administrators can open it, for support and abuse prevention.
Application texts contain personal data. Check them only where your process and privacy notice allow it, and do not submit more of a candidate's documents than you need.
Check a document or web page for AIResponsible use of Textumi
Questions fréquentes
Can an AI detector tell if a cover letter was written with ChatGPT?
No. A detector only estimates how close the wording is to AI writing; it cannot see how the letter was made or name a model. Cover letters tend to be formulaic and heavily polished, so human letters are often flagged as well. Take a result as a reason to ask a question, but never as proof.
Is it fair to reject a candidate because a detector flagged their application?
No. A detector result is a statistical signal about wording, and it is wrong in both directions, especially on short, formal or second-language texts. A person should decide on the substance of the application and on what the candidate shows in the process, not on a label.
Does the EU AI Act apply to AI detectors in recruitment?
In its Annex III, the AI Act lists AI systems intended to analyse and filter job applications or assess candidates among high-risk systems. Whether this applies to a specific use depends on how the tool is used in your process. Before you use any tool to filter applications, speak to your data protection officer or a lawyer.
Should we ban AI-written cover letters?
That's up to you. It's hard to enforce and easy to get wrong, because detectors can't prove authorship. Instead, many employers state that they're fine with AI help as long as everything in the letter is true. Interviews and work samples then show what matters: whether the person can do the job.
What should a candidate do if they are told their letter looks AI-generated?
Stay calm and explain how you wrote it. Show your drafts or notes if you have them, and talk about the experience the letter describes in your own words. A specific, honest answer about your work shows authorship far better than any detector result.
Is a candidate's application safe to check in Textumi?
Submitted texts and results are stored encrypted, and checks use Textumi's own models. On Individual and Professional, checks stay in a private History for 30 days unless deleted earlier. On all other plans you can open a result for about an hour after it finishes; after that it is no longer available to you, and it is deleted within 7 days. Until then it is kept encrypted, and only Textumi's administrators can open it, for support and abuse prevention. Check applications only in accordance with your privacy notice.