Skip to content
Start free
Platform guide

Last verified

Does Turnitin detect AI? (October 2026)

Yes. Turnitin has its own AI writing detector, but only institutions that license Turnitin Originality get it. It reads submissions of 300 to 30,000 words of long-form prose and shows instructors the share of text likely written by AI. Turnitin says the score can be wrong and should never be the sole basis for action.

Check your text freeCreate a free account

Quick answer

Detects AI on its own
Yes, built in
Detector it uses
Turnitin AI writing detection
What it is
Plagiarism and integrity checker

Yes, when the school licenses Turnitin Originality. It needs 300 to 30,000 words of prose and reports to instructors, not students, by default.

Turnitin's own help page (opens in a new tab)

Checked against Turnitin's documentation on .

Check your text before it's submitted

Whatever Turnitin runs, you can see how a draft reads to an AI detector first. Paste it in for an overall score and a highlight on every sentence that reads as machine-written.

A score is evidence, not proof. Read the highlighted sentences before you decide anything.

  • 99.4% accurate
  • Free to try
  • 14 languages
  • Sentence-level results
0 words

Try a sample

How Turnitin checks for AI

What Turnitin documents, checked . Schools choose their own settings, so yours may differ.

  • It only runs where the school pays for it

    AI writing detection is part of Turnitin Originality. Institutions on Turnitin Similarity or Feedback Studio alone don't get it, and if the main account doesn't license Originality, none of its sub-accounts do either.

    Source: helpcenter.turnitin.com (opens in a new tab)

  • It needs enough of the right kind of text

    A submission must have at least 300 words of prose in a long-form format and no more than 30,000 words, be under 100 MB, and arrive as .docx, .pdf, .txt or .rtf. Turnitin says the model doesn't reliably detect AI in poetry, scripts, code, bullet points, tables or annotated bibliographies.

    Source: guides.turnitin.com (opens in a new tab)

  • Low scores are hidden on purpose

    Turnitin found false positives are more common below 20%, so scores from 1% to 19% appear as *% with no percentage and no highlights. Reports made before July 8, 2024 may still show a number in that range.

    Source: guides.turnitin.com (opens in a new tab)

  • Paraphrased and bypassed text: English only

    The report breaks the score into AI-generated text and AI-generated text that was then run through a paraphraser such as QuillBot. Since August 2025 the AI-generated category also covers text likely changed by a bypasser (humanizer) tool. Turnitin says these paraphrase and bypasser checks only work on English submissions.

    Source: guides.turnitin.com (opens in a new tab)

What instructors see

  • The AI score is separate from the Similarity score

    Turnitin's AI percentage is independent of the similarity score, and AI highlights don't appear in the Similarity Report. Open the AI Writing tab (or the AI indicator in the classic report) to see it.

  • An asterisk is not a zero

    *% means Turnitin found some AI-like text but below the 20% threshold where it reports a number. It is a deliberate low-confidence state, not a finding.

  • Use it as one piece of evidence

    Turnitin says its model may misidentify human-written, AI-generated and AI-paraphrased text, and that deciding on misconduct takes human judgment and your institution's own policy.

  • Grammar tools are treated differently

    Turnitin says ordinary Grammarly-style spelling and grammar fixes were mostly not flagged in its tests, but text produced by generative features such as drafting or paraphrasing likely will be.

What students should know

  • You may never see the AI score

    Turnitin's FAQ says the AI indicator and report are for instructors and administrators. An instructor can download the report and share it, and some setups let instructors make it available to students.

  • Short and non-prose work is a weak signal

    Under 300 words of prose, Turnitin produces no AI score at all, and lists, tables, code and poetry aren't reliably assessed. If a flag lands on that kind of work, say so.

  • Have your process ready

    If you're questioned, drafts, notes and version history carry more weight than any score. See our guide on how to prove you didn't use AI.

Flagged for AI when you wrote the work yourself? Here's how to show your work.

Turnitin is the detector most schools mean when they talk about AI checks. Whether it runs on your work depends on your institution's licence, the length and type of what you submit, and the language it's written in. Here is what Turnitin's own documentation says about each.

What the percentage means

Turnitin's overall percentage is the share of qualifying text, meaning prose sentences in long-form writing, that its model judges was likely generated by AI or generated and then altered with a paraphraser or bypasser. It is not a count of how many words a chatbot typed, and a document that mixes prose with lists or tables can show a percentage that doesn't match the highlights.

Languages it reads

Turnitin's AI detection covers English and several other languages, including Spanish, Japanese and Arabic, each with its own model. A submission in an unsupported language gets no AI report. The extra checks for paraphrased and bypassed text run on English only.

Checking your own draft

Students can't run Turnitin's AI check on their own. To see how a draft reads first, paste it into the scan box on this page: Winston AI gives a 0 to 100 human score and highlights sentence by sentence what reads as AI. Free checks take 80 to 800 words, no account needed. A different detector can score the same text differently, so treat any result as a prompt to revise, not a prediction.

Questions about Turnitin and AI detection

Still unsure? Browse our guides or ask our team.

Does Turnitin detect AI?

Yes, if your institution licenses Turnitin Originality. Its AI writing detection estimates how much of a submission's long-form prose was likely written by AI and shows the result to instructors.

Does Turnitin detect ChatGPT?

It is built to catch text from large language models, including ChatGPT. It reports a likelihood for each passage, not the name of the tool used, and Turnitin warns that it can misidentify human writing.

What AI detector does Turnitin use?

Its own. Turnitin built and trains its AI writing detection model, with separate models for languages such as Spanish and Japanese, and updates it several times a year.

How does the Turnitin AI detector handle short essays?

It doesn't score them. A submission needs at least 300 words of long-form prose, and Turnitin says lists, tables, code and poetry aren't reliably assessed.

Is the Turnitin AI checker available to students?

Not as a self-check. The AI indicator is meant for instructors and administrators, who can download the report and share it with you.

Can Turnitin detect QuillBot, paraphrased or humanized text?

For English submissions, Turnitin says it can flag AI text that was run through a paraphraser like QuillBot or a bypasser (humanizer) tool. Those checks don't run on Spanish, Japanese or other languages, and like every detector they can be wrong.

What does *% mean on a Turnitin AI report?

It means a low score that Turnitin deliberately hides. Results between 1% and 19% are shown as an asterisk with no highlights, because false positives are more common in that range.

Does Grammarly get flagged as AI by Turnitin?

Usually not for grammar fixes. Turnitin says spelling and grammar corrections from tools like Grammarly were mostly not flagged in its tests, but text written or paraphrased by Grammarly's generative features likely will be.

Sources

Last checked

Other platforms

See what a detector sees

Paste any draft to get an overall score and sentence-level highlights. A quick scan doesn't need an account.

Scan text freeSee plans