Detectize

How it works

  1. 1

    Stay calm

    A detector score is an estimate, not proof of who wrote a text.

  2. 2

    Collect your evidence

    Drafts, version history, notes and sources show how your work was made.

  3. 3

    Ask for a conversation

    Walk your instructor through your process instead of arguing about the number.

Why we write about AI detectors

Schools and teachers use tools like Turnitin, GPTZero and the Grammarly AI checker (often called the Grammarly AI detector) to flag AI-written work. The problem: these checkers sometimes flag text that a student wrote entirely by themselves. Detectize explains how these scores work, what to do when your own writing is flagged, and what we found when we tested a detector ourselves.

What the Grammarly AI detector score means

Many students search for the Grammarly AI detector because they want to know how their essay will look to an AI checker before a teacher sees it. Its percentage, like every detector's, is an estimate of how predictable the wording is, not a count of AI-written words. We explain how to read it in what the Grammarly AI detector percentage means, and what to do next in Grammarly says my essay is AI but I wrote it.

Different detectors are trained differently, so they often disagree on the same text. A low score in one tool doesn't guarantee a low score in another, and a high score isn't proof of anything. The most useful way to treat any checker is as an early warning: find the sentences that sound generic, rewrite them in your own words with specific details, and keep your drafts so you can show how your work was made.

How accurate are AI detectors?

Less accurate than their scores make them look. Every AI detector gives you a probability, not a verdict. Even good ones make mistakes in both directions: they miss some AI-written text, and they flag some text that a person wrote.

We didn't want to just repeat that warning, so we ran a small test of our own. We checked 29 passages of known origin through a dedicated AI-detection engine: 15 written by people and 14 written by an AI assistant. It flagged 13 of the 14 AI passages as likely AI. It also flagged 6 of the 15 human passages. The human passages it got wrong were mostly modern, plain-language writing, such as public-health and science pages from U.S. government websites. Older literary prose was almost never flagged.

This is a small sample, not a benchmark, and we only tested one detection engine. It does not tell you how often a false flag will happen to your own work. It does show that clear, neutral, well-organized writing, which is exactly what students are taught to produce, can be flagged as AI.

The full results, including a comparison with TypeSafe's Jev model, are in our AI detector test.

Why clear writing can look "AI-written" to a detector

AI detectors don't know who wrote a text. They estimate how predictable the wording is. Language models tend to choose likely words in a smooth, even rhythm, so text that is predictable, evenly paced and full of standard phrasing scores as more AI-like. A careful explanation of a technical topic often looks like that, whether a person or a model wrote it.

That is why the same detector can give one paragraph of a report 1% and the next paragraph 99%. The score reflects how the sentences read, not where they came from.

What to do with a detector result

Treat the score as a second opinion, not a judgment. Look at which sentences are highlighted and ask whether they sound generic: stock transitions, vague claims, or sentences that could appear in any essay on the topic. Replacing them with specific examples, your own observations and your own way of explaining things usually helps, and it makes the writing better anyway.

Keep your drafts and version history while you work. If someone ever questions your writing, the record of how it was made is stronger evidence than any detector score, in either direction.

Frequently asked questions

Is Detectize the official Grammarly AI detector?

No. Detectize is an independent site with guides and tests about AI detectors. We are not affiliated with Grammarly and we don't run a checker on this site at the moment.

Best software to detect cheating on essays?

No software can prove that an essay involved cheating. AI detectors like Turnitin or GPTZero estimate how likely a text is to be AI-written, and they make mistakes in both directions, including flagging text people wrote themselves. Treat a high score as a reason to talk with the writer and look at drafts, notes and version history, not as evidence on its own.

Do you use cookies?

We count visits with our own cookie-free statistics, and only if you click "Accept all" do we set one statistics cookie that remembers how you first found the site. You can change your choice any time under "Cookie settings" in the footer.

Why does my human-written essay show as AI?

AI detectors look for predictable wording and uniform sentence length. Formal, polished or non-native English can trigger false positives. Our guide on what to do if you're flagged walks through the next steps.

How accurate is an AI detector?

No AI detector is 100% accurate – including the ones schools use. In our own small test, the detector we used caught 13 of 14 AI passages but also flagged 6 of 15 passages written by people. Treat any score as an estimate, not proof.

Can an AI detector flag text I wrote myself?

Yes. In our own small test, the detector we used flagged 6 of 15 passages that people wrote, mostly plain modern prose. A flag is a reason to review your text, not proof that you used AI.

Is the Grammarly AI detector accurate?

We haven't tested Grammarly's detector, so we can't give you an accuracy number for it. What we can say from our own testing is that no detector is accurate enough for its score to be treated as proof.

Do different AI detectors give the same score?

Not necessarily. Detectors are trained differently and often disagree on the same text. In our test, one general model and one dedicated detector gave very different answers on the same plain government science pages. A low score in one tool says little about another.

AI Detection Guides

Flagged by an AI detector?

Start with the guide that fits your situation: what to do now, how to prove your process, or what our own test showed.

See the guides

Browse Guides · About Detectize · Contact