The toolkit / Résumé
Free AI detector for résumés and cover letters
See how much of your writing reads as model output, and what to change.
The overview
A little less work. A better starting point.
Paste any text, or drop in a résumé, and get a percentage for how strongly it reads as default AI writing, along with the exact words and sentence habits that gave it away. Rule-based and offline: no model is called, so nothing you write is sent anywhere.
Try it here
Run the check on this page.
The same scoring Vulture uses, inside this tab: nothing is uploaded. The full report is one click away in Vulture.
01 / Capabilities
What it does
Evidence, not a verdict
Every signal carries the snippets that triggered it: the vocabulary that spiked after 2022, stock phrases, tricolons, negative parallelism, participial tails, uniform sentence length and em-dash habits.
Word-by-word replacements
Flagged words come with plainer alternatives. Delve becomes look into, leverage becomes use, robust becomes reliable, and the sentence stops announcing itself.
Human markers count too
Contractions, specifics, numbers and irregular rhythm push the score down. The check measures what is missing as well as what is present.
Model fingerprints
Some phrasings belong to particular model families. Where a fingerprint is strong enough to be worth naming, it is named, with the lines that carry it.
Runs offline
The whole check is a rule set, not an API call. It works with no network, no key and no account, and your text is never transmitted.
02 / The workflow
How to use it
- Open the free check and drop in a résumé, or paste any text you want to test.
- Read the AI meter: low is good here. Above 65 means the writing reads as generated.
- Open the flagged lines and swap the marked words for the plain alternatives offered.
- Rewrite the sentences that all have the same length; irregular rhythm is the strongest human signal.
- Check again. The score should fall as the specifics go up.
03 / Good to know
Questions, answered.
Can an AI detector be trusted?
No detector can prove authorship, and this one does not claim to. It measures how closely a text matches documented model habits, and it shows you every reason for the number so you can judge it yourself.
Will a human-written résumé score high?
It can. Résumés are formulaic by nature: parallel bullets, no contractions, action verbs throughout. Read the signals rather than the number alone, and treat the flagged lines as writing advice.
Is my text sent to a model?
Never. The check is a rule set running in your browser. There is no API call, so there is nothing to intercept and nothing stored.
Does it work on cover letters and emails?
Yes. Text can be checked as a résumé, an email or general prose, and each mode weighs the signals differently because the norms differ.
Better together
Use it with
More free tools