Home / Guides

AI detector for hiring?

Cover letters are some of the worst inputs you can give a detector: short, formulaic, and full of standard phrasing. Here is what a score actually means on hiring documents, and how to run a fair process if you check anyway.

Updated October 31, 2026 · By the IsThisAI? team

The short answer

AI detectors can flag an AI-written cover letter, but they cannot tell you who wrote anything, and they are wrong often enough that no hiring decision should rest on a score alone. If you use one, treat it as a prompt to look closer, never as evidence.

Why hiring is a hard case for detectors

Cover letters and resumes are some of the worst inputs you can give a detector. They are short, formulaic, and full of standard phrasing. "I am excited to apply for" and "detail-oriented team player" appear in millions of human-written applications, and they also appear in millions of AI-written ones. A detector looking for statistical fingerprints finds very little to grip.

Research on detector accuracy consistently shows the same pattern: short text is where detectors fail most. A cover letter of 200 words gives the model far less signal than a 2,000-word essay. False positives cluster exactly where you do not want them: polished, formal writing, and writing by non-native English speakers, whose simpler sentence structures can read as "AI-like" to a model trained mostly on native writing.

What a score actually means on a resume

Most resumes are bullet points and fragments, not sentences. Detectors are built for prose, so scores on resumes are close to meaningless. A 90% score on a resume does not mean a candidate used AI. It usually means the resume is written in the standard compressed style that every career guide teaches.

Cover letters fare slightly better because they contain connected prose, but the same caution applies. An 80% score is a reason to read the letter carefully, not a reason to reject the candidate.

A fair process if you check anyway

If your team decides to screen with a detector, build the process around its limits:

• Never auto-reject on a score. Use a high score only as a trigger for human review.
• Check the work sample, not just the letter. Longer, more specific writing gives detectors more to work with and gives you more to judge.
• Ask, do not accuse. "Walk me through how you put this together" tells you more than any score. Candidates who did the work can talk about it in detail.
• Apply the same check to everyone. Spot-checking only some candidates, especially ones whose names or backgrounds differ from yours, is a bias lawsuit waiting to happen.
• Write the policy down. If AI use in applications is against your rules, say so in the job posting. If it is allowed, say that too. Candidates deserve to know the rules before they play.

The question underneath

Ask what you are actually trying to learn. If the concern is that a candidate cannot write, a short paid work trial or a writing exercise in the interview answers it directly. If the concern is honesty, note that using AI to polish a cover letter is now as normal as using spellcheck, and many strong candidates do it openly. A detector cannot tell you which kind of candidate you are looking at. Only the interview can.

Bottom line

Detectors are a weak signal on hiring documents. Use them, if at all, as one input among many, with a written policy, applied evenly, and never as the deciding factor. The candidates worth hiring will show it in the conversation.

What candidates are actually doing with AI

It helps to know what you are screening for. Most candidates who use AI on applications are not generating entire cover letters from a blank prompt. The common patterns are polishing (fixing grammar and flow in a letter they wrote), structuring (asking for an outline, then writing it themselves), and tailoring (feeding the job description in and asking what to emphasize). All three leave different traces, and the last two are arguably just good preparation.

This matters because the line has moved. A few years ago, using any tool beyond spellcheck felt like cheating to many hiring managers. Today, a large share of candidates use AI assistance openly, the way they use LinkedIn or a resume template. Treating every AI-assisted letter as dishonest will filter out strong candidates who are simply using normal modern tools. Decide what your actual rule is: is the problem AI use, or is the problem a candidate who cannot do the job? Those are different problems with different tests.

Better signals than a detector score

If the underlying worry is competence, there are direct ways to check it that do not involve detectors at all:

• A short work sample. Give finalists a small, paid task that mirrors the job. Their output tells you what they can do, no detector needed.
• Structured interviews. Ask every candidate the same questions and score the answers the same way. "Tell me about a time you..." reveals depth that a cover letter never will.
• Live writing. For writing-heavy roles, a 20-minute exercise during the interview shows real ability far better than a polished letter of unknown origin.
• References who saw the work. A former manager describing what the candidate actually produced beats any document screen.

None of these can be gamed the way a cover letter can, and none of them risk punishing a good candidate for writing cleanly.

Sources

1. Weber-Wulff et al., "Testing of detection tools for AI-generated text," International Journal for Educational Integrity 19, 26 (2023). https://link.springer.com/article/10.1007/s40979-023-00146-z
2. Liang, Yuksekgonul, Mao, Wu & Zou, "GPT detectors are biased against non-native English writers," Patterns 4(7), 2023. https://www.cell.com/patterns/fulltext/S2666-3899(23)00130-7
3. Turnitin, "Turnitin's AI writing detection capabilities FAQs." https://www.turnitin.com/products/features/ai-writing-detection/faq

This guide is general information, not legal or hiring advice. AI detection results, including ours, can be wrong.