AI Detector for Teachers

Check a student's essay sentence by sentence. A flag is a reason to talk, not proof.

0 / 2,000 words

How to check a student's essay for AI

  1. Step 1

    Paste or upload

    Paste the essay, or sign in to upload a .docx, .pdf or .txt file. You need at least 150 words (200 in Spanish, Portuguese and French) for a result.
  2. Step 2

    Look at the colors

    Red sentences read like AI, yellow ones are unclear, and grey text wasn't scored.
  3. Step 3

    Read the verdict carefully

    The overall result is likely AI, mixed or likely human. Mixed means there's no confident verdict on the text as a whole.

What the report shows

  • An overall AI score from 0 to 100 and a verdict: likely AI, mixed, or likely written by a person.
  • Sentence colors, so you can see where the score comes from instead of taking one number on trust.
  • A note when part of the text is in a language we don't score yet, so you know what was and wasn't checked.

The report doesn't say which AI tool wrote something. It only estimates how much the writing reads like AI. It's the same report our main AI writing detector gives; this page adds the guidance for teachers.

“Mixed” means there's no confident verdict on the essay as a whole. It doesn't mean part of it was written by AI. A mixed essay is worth a closer look at the red sentences, nothing more.

If an essay mixes languages, for example an English essay with quotes in Spanish, the result says which part the verdict covers and how many words were scored. Grey text wasn't checked at all.

False positives and English learners

A false positive is human writing flagged as AI. These are our results on English texts the detector never saw while we tuned it. The last column is the 95% confidence bound: how far the real rate could plausibly be from what we measured.

False positives and detection on our English test set
TextsHow manyResult95% bound
Human-written texts1,0961 flagged as likely AI (0.09%)At most 0.52%
Essays by non-native English writers6000 flagged as likely AIAt most 0.64%
AI-generated texts from 6 model families278270 caught as likely AI (97.1%)At least 94.4%
English, November 2026, detection engine version w0p5+es+fr+pt-1a15681b+avg3frpt.

Be cautious with short pieces, essays written to a strict template, and careful writing by English learners. Our test results don't guarantee a correct verdict on an individual essay.

If a student tells you their own work was flagged, our guide on what to do after an AI detector false positive is written for them and worth passing on.

Other signs to check before you decide

  • Version history: documents written over days look different from text pasted in all at once.
  • Drafts and notes: ask to see outlines, earlier versions or research notes.
  • Writing done in class: compare the essay's voice and level with work you've seen the student write.
  • Sources: check that quotes and references exist and say what the essay claims they say.
  • The student's own explanation: can they talk through their argument and how they got there?

Check references that look unfamiliar: search for the title and the author. A journal article that doesn't exist, or one that exists but says something else, is worth raising with the student whatever the AI score says.

It also helps to look at the essay's shape. A student who usually writes long, uneven paragraphs and suddenly hands in five tidy paragraphs of the same length may simply have worked harder, but it's a fair thing to ask about.

Using the detector fairly across a class

The way you use a detector matters as much as the score. If you only check the essays that already feel off, every flag will seem to confirm your hunch, and students who write in a plain or formal style get checked more often than everyone else.

  • Tell students at the start of the term whether you check work for AI, and what happens if something is flagged.
  • Check every essay for an assignment, or none of them. Don't pick out a few.
  • Write down the date, the score and the sentences that were flagged, so you can explain later what you saw.
  • Set expectations for tools in writing: whether grammar checkers, translators or AI brainstorming are allowed, and whether students should say when they used them.

If many essays in one class come back mixed or likely AI, look at the assignment before you look at the students. Prompts that ask for a summary of a well-known topic, or a fixed five-paragraph structure, tend to produce writing that reads alike, whoever wrote it. Asking for personal examples, class discussions or local sources makes the work harder to hand off to a chatbot and easier for you to judge.

Talking to a student about an AI flag

If you decide to raise it, give the student time to explain how they wrote the essay.

  1. Start by asking how they wrote the piece, before mentioning the detector.
  2. Ask them to bring drafts, notes or the document's version history.
  3. Go through one or two flagged passages together and ask them to explain the reasoning.
  4. Don't decide anything in the first conversation.
  5. Follow your school's academic integrity process for anything beyond that.

Questions that tend to open the conversation without accusing anyone:

  • “Walk me through how you started this essay.”
  • “Which source was most useful, and where did you find it?”
  • “This paragraph makes a strong claim. How would you explain it in your own words?”
  • “Did you use any tools while writing, like a grammar checker or a translator?”

Our academic integrity page explains how we expect people to use AI tools and detection scores.

Which detectors schools use, and where this one fits

Many schools use the AI detection built into Turnitin through their learning management system. This tool is different: it's for individual teachers who want to check a piece of writing quickly and see sentence-level detail. We checked 32 colleges' policies in what AI detectors colleges use.

There's no LMS integration and no class account yet. You paste or upload one piece at a time.

If your school already uses another detector, its score and ours will often differ, because each tool is built and tuned differently. If the two disagree, look at the student's drafts and the flagged passages before you draw any conclusion.

To compare detectors on false positives and price, see our roundup of the best AI detectors.

Free limits for teachers

AI detector limits for teachers
Checks a dayWords per check
No account102,000
Free account3010,000
Paid plan20010,000
There's no school plan or batch checking yet. Each check is one text.

Uploading .docx, .pdf and .txt files needs a free account. After an upload you see the extracted text before it's checked, so you can remove a cover page or a reference list you don't want included.

Student privacy: what happens to the text you paste

  • Without an account, the text isn't saved on our servers. Your browser keeps a temporary copy so it survives a refresh.
  • With an account, the text is saved as a draft in your history for 7 days on a free plan, or 30 days if you had a pass or subscription when it was saved. You can delete it any time.
  • Only you can see your history. We never use the text to train models.

The full details are in our privacy policy.

Frequently asked questions

Is this AI detector free for teachers?

Yes. You can run 10 checks a day without an account, or 30 a day of up to 10,000 words each with a free account.

How accurate is it on student essays?

On our November 2026 English test set it flagged 1 of 1,096 human texts and none of 600 essays by non-native English writers, and caught 270 of 278 AI texts. It can still be wrong; review drafts and discuss flagged passages with the student.

Can it detect ChatGPT, Claude and Gemini?

Yes. It reads how the writing sounds, not which tool produced it, and our test set includes text from GPT, Claude, Gemini and three other model families.

Can students get around it with a humanizer or paraphraser?

Sometimes, yes. Text that has been heavily rewritten is harder for any detector to catch. Turnitin added detection of AI bypasser tools for English text in August 2025. Our published accuracy figures cover unedited AI text.

Should I fail a student based on an AI score?

No. The score estimates how likely the text is to be AI-written. Review drafts, talk with the student and follow your school's process.

Does it work on essays by English learners?

In our test, none of 600 essays by non-native English writers were flagged as likely AI. Short, formulaic or very careful writing can still score higher, so be especially cautious with these essays.

Can I check Spanish or Portuguese essays?

Yes, both.

Do you keep my students' essays?

Without an account, we don't save the text on our servers. With an account, it's saved as a draft in your history (7 days on a free plan) and you can delete it at any time. We never use it to train models.

Turnitin is a trademark of its owner. AI Checker Humanizer isn't affiliated with Turnitin.

Try it on your own text

Check for AI