Updated August 2026 · read from Turnitin's own FAQ
Does Turnitin detect AI? Does Turnitin check for AI writing, and what the Turnitin AI detection tool actually checks
Not on every paper, and not at every school. Turnitin publishes three conditions that all have to be true before a single sentence of your work is scored for AI.
Below: the three gates, what the detector reads and what it skips entirely, and what the percentage means once your instructor has it.
2 free runs a day, up to 200 words each. We save your run so you can get back to it, and a delete button appears with the result. See privacy.
This is our own AI-pattern score, measured here on sentence rhythm, template phrases, vocabulary variety and passive voice. It is not a GPTZero, Turnitin, Originality.ai, Copyleaks or ZeroGPT result, and it does not predict one. Worth knowing: we also ask the rewrite to vary sentence length, drop template phrases and prefer the active voice, so some of the drop is built in. Read the two panels below, not just the number.
Before ·
After ·
Your own text rewritten · Saved to your history, delete any time
The short answer
Turnitin detects AI writing, but it does not check every submission. Three conditions have to hold. The institution has to license Turnitin Originality, because AI writing detection "is only available to customers that license Turnitin Originality" and is not included with Turnitin Similarity or Feedback Studio on their own. An administrator has to leave the feature enabled, because Turnitin "will only process submissions for AI writing detection if the institution has the feature enabled" and admins can switch it off or opt out completely. And the submission itself has to qualify: at least 300 words of long-form prose, under 30,000 words, in English, Spanish or Japanese. Where all three hold, every submission is scored automatically. Where any one fails, nothing is scored at all, and you have no way to tell which case you are in, because the indicator is hidden from students either way.
Last updated August 15, 2026. Every Turnitin statement quoted here was read from turnitin.com on that date.
The conditional answer
The three gates between your paper and an AI score
Most pages answering this question treat Turnitin AI detection as a single switch that is always on. Turnitin's own documentation describes a licensing decision, an administrator setting and a file requirement, in that order.
| Condition | What Turnitin says | If it is not met |
|---|---|---|
| 1. Your institution licenses Turnitin Originality | AI writing detection is not part of every Turnitin product. Turnitin states it "is only available to customers that license Turnitin Originality" (or the AI writing add-on for iThenticate 2.0). Institutions running Turnitin Similarity, Turnitin Feedback Studio or Originality Check on their own do not have it. | Nothing is checked for AI. The Similarity Report has no AI indicator in it at all. |
| 2. An administrator has the feature switched on | Licensing it is not the same as running it. Turnitin: "admins have the option to enable/disable the AI writing feature from their admin settings page." Institutions can also opt out of processing entirely, "by disabling AI writing detection from their admin account settings page". | Turnitin says plainly that it "will only process submissions for AI writing detection if the institution has the feature enabled". Your paper is never scored. |
| 3. Your submission qualifies | The model reads long-form prose only, needs at least 300 words of it, stops at 30,000 words, and runs on English, Spanish and Japanese. Files must be .docx, .pdf, .txt or .rtf under 100 MB. | The submission processes but returns no score, or returns a percentage covering only the qualifying part of the document. |
The licensing gate is the one almost nobody writes about, and it is the reason two students at different universities can get completely different answers to the same question and both be right. Turnitin sells several products. Similarity matching is the old, universal one. AI writing detection arrived later as part of Turnitin Originality, and Turnitin's FAQ tells Feedback Studio and Similarity customers to "speak to your Turnitin account manager regarding access". An institution can be a long-standing Turnitin customer and still have no AI detection anywhere in its workflow.
The administrator gate is separate again. Licensing the product does not force it on. Turnitin gives admins an explicit toggle, and disabling it "will remove the AI writing indicator and report from the Similarity report" for instructors and admins alike. Some universities enabled it in 2023, watched the false-positive coverage, and turned it back off. Others never enabled it at all. This is a policy decision made above your head and rarely announced to students.
There is one integration quirk worth knowing if your course runs through Microsoft Teams. Turnitin says AI writing detection is available with the Feedback Studio integration, but not with the Teams Assignment Similarity integration, which uses only the student viewer. In that setup an instructor who has a concern has to ask an administrator to run a full report separately.
Canvas adds a gate of its own on top of these three. Turnitin is not a campus-wide filter there: an instructor has to enable it on each individual assignment before anything is sent, so two papers in the same course can be treated completely differently. We work through all four conditions, and where the similarity score fits in, on our Canvas AI detector page.
Coverage
What Turnitin checks for AI, and what it skips
Even with everything enabled, the detector reads a narrower slice of your document than most people assume. It looks for prose sentences inside paragraphs and ignores the rest before it calculates anything.
| Content | Checked for AI? | Why |
|---|---|---|
| Essay, dissertation, article, report body | Yes | Turnitin defines qualifying text as "individual sentences contained in paragraphs that make up a longer piece of written work". |
| Bullet points and short non-sentence structures | No | Named directly as short-form or unconventional writing the model does not reliably detect. |
| Tables and annotated bibliographies | No | Excluded for the same reason. They are not prose sentences inside paragraphs. |
| Code | No | Turnitin: "we are not pursuing ChatGPT code detection at this time." |
| Poetry and scripts | No | Non-prose. The model is not built to read it. |
| Anything under 300 words | No | A hard file requirement, not a soft threshold. Short reflections and discussion posts fall below it. |
| Handwritten work in Gradescope | No | Turnitin says it has no current plans to add AI detection there, because Gradescope handles handwriting and AI detection targets typed text. |
| A paper submitted before the feature was turned on | No | Turnitin: "we cannot retroactively process submissions. If you would like to process past submissions for AI writing, you will need to re-submit the document." |
This is why the highlighting and the percentage so often look like they disagree. If a third of your dissertation is tables, figure captions and a bibliography, none of that reaches the model. The percentage is calculated over the qualifying prose alone, so a figure that looks alarmingly high may be describing a much smaller body of text than the page count suggests.
It also explains a class of submission that is simply never scored. Discussion board posts, short reflections, lab summaries and problem sets rarely clear 300 words of continuous prose. Turnitin treats that as a file requirement rather than a preference, so those submissions come back with no AI reading at all, whatever the institution has enabled.
The second layer
Turnitin also checks for paraphrasing and bypassing tools
The question people usually mean by "does Turnitin check for AI" is narrower than the one they ask. They want to know whether running an AI draft through a rewriter changes the answer. Turnitin addresses that directly, and the detail is more specific than the general accuracy claims around it.
Turnitin says its capabilities include "detection of likely AI-generated content that may have been modified using a word spinner/AI paraphrasing or bypassing tool to evade detection", and that this "is run automatically for all submissions" wherever the AI writing indicator is enabled. It is not a separate report or an extra setting. It is folded into the same percentage, so a submission is checked for AI writing, AI bypassing and AI paraphrasing in one pass.
Two limits are worth having straight. The paraphrasing and bypassing layer is English only, narrower than the main detector's English, Spanish and Japanese. And Turnitin refuses to say which tools it catches: asked for a list, it answers that it is "unable to disclose the names of these tools" because sharing one "would make it easier for students to evade our system". Any product promising you a guaranteed clear Turnitin score is promising something the vendor of the detector has deliberately made unverifiable.
What that leaves is the writing itself. Turnitin names the prose most likely to be misread, and one of the three categories is "text that has been paraphrased without developing new ideas". Synonym-swapping keeps the flat rhythm and the recycled structure that the model responds to, which is why a lightly reworded draft often reads worse to a detector rather than better. We work through that in detail in whether Turnitin detects QuillBot paraphrasing.
The output
What the Turnitin AI detection tool actually reports
Six things about the percentage that change how much weight it deserves, all of them published by Turnitin rather than inferred from testing.
One percentage, and it is not a share of your document
The number covers qualifying text only. Turnitin warns that it "does not necessarily correlate to the amount of text in the submission", because everything that is not long-form prose is excluded from the calculation before the percentage is worked out. A paper that is half tables can show a figure drawn from the other half.
Nothing at all between 1% and 19%
Turnitin attributes "no score or highlights" in that band and displays an asterisk instead. Its reason is its own testing, which "found that there is a higher incidence of false positives when the percentage is between 0 and 19". The instructor does not get a number there either.
A deliberate under-count
Turnitin publishes what it trades away to keep false positives down: "if we identify that 50% of a document is likely written by an AI tool, it could contain as much as 65% AI writing." The reported score is a floor rather than an estimate.
A separate measurement from similarity
The AI percentage is "different from and independent of the similarity score", and AI highlights never appear in the Similarity Report itself. A clean similarity score says nothing about the AI score, and reading the two as one number is the most common mistake students make.
Visible to your instructor, not to you
Turnitin: "only instructors and administrators are able to see the indicator", and "the AI writing detection indicator and report are not visible to students". Instructors can download the AI report as a PDF and share it, which some do and most do not.
Not a finding of misconduct
Turnitin writes that the tool "does not make a determination of misconduct" and that the percentage "should not be used as the sole basis for action or a definitive grading measure by instructors". How closely a given institution follows that instruction is where the real risk sits.
Turnitin validates the detector against "over 700,000 additional academic papers that were written before the release of ChatGPT" and re-tests before every model release. It targets false positives "under 1% for documents with over 20% of AI writing", a figure with a condition attached that is easy to miss: the rate below 20% is not published, and that is precisely the range a wrongly flagged student lands in. We trace every accuracy number in circulation, including the several that have no traceable source, in how accurate the Turnitin AI detector is.
If you want the mechanism rather than the policy, the model extracts sentences, segments them into overlapping sections, scores each segment between 0 and 1, lets qualifying sentences inherit their segment score, then pools the overlaps and aggregates. That process is unpacked step by step in how Turnitin detects AI writing.
The practical part
What to do when you cannot tell whether it is on
The honest position is that you will usually not know. The indicator is hidden from students whether the feature is enabled or disabled, so the two cases look identical from your side. That makes the useful moves the ones that hold up either way.
Read your integrity policy, not the forums. Whether AI detection is licensed and enabled is a decision your institution made and usually wrote down. The course syllabus, the academic integrity policy and the library's guidance are far more reliable than a Reddit thread about a different university.
Keep evidence of how you wrote it. A version history that shows a document developing over days is worth more than any score, because it addresses the actual question in a false-flag conversation. Draft in a tool that records revisions and avoid pasting a finished draft into an empty file. We cover what stands up and what does not in how to prove you did not use AI.
Assume it is on and write accordingly. Given the cost of guessing wrong, the sensible default is to treat every submission as scored. That does not mean gaming a number. It means the prose should not read like the flat, evenly-paced, idea-restating writing that Turnitin itself names as the most likely to be misread.
That last point is the work our humanizer does. It rewrites uniform AI drafting into prose with real variation in sentence length, structure and register, which is what separates writing a person produced from writing a model produced. It cannot tell you whether your school has the feature switched on, and it cannot promise you a number, because Turnitin has made both of those unknowable on purpose.
Frequently asked
Turnitin AI detection questions, answered
Does Turnitin check for AI?
Only when the institution licenses Turnitin Originality and an administrator has enabled the feature. Turnitin states it "will only process submissions for AI writing detection if the institution has the feature enabled". Where those conditions are met it runs automatically on every submission, alongside the similarity check.
Does Turnitin detect AI?
Yes, where it is enabled. It scores submitted text for the likelihood that a large language model wrote or modified it, covering the GPT-5 and GPT-4o families, Claude, Gemini, Grok, LLaMA, Mistral and DeepSeek among others. It works from the finished document alone and cannot see your prompts or your account.
How do I know if my school has Turnitin AI detection turned on?
You usually cannot tell from your own view, because the indicator is hidden from students whether or not it is enabled. Your course or academic integrity policy is the reliable place to look, and your instructor or library can answer directly. Assume it is on unless you have been told otherwise.
Can my institution turn Turnitin AI detection off?
Yes. Turnitin: "admins have the option to enable/disable the AI writing feature from their admin settings page", and an institution that does not want submissions processed at all "can opt out by disabling AI writing detection from their admin account settings page". Several universities have done exactly that.
What does Turnitin detect besides AI?
Its original job is text matching. The Similarity Report compares your submission against web pages, publications and a repository of student papers, and reports the percentage that matches something else. That runs on every submission. AI writing detection is a separate feature layered on top of it.
Does Turnitin check old submissions for AI?
No. Turnitin is explicit: "we cannot retroactively process submissions. If you would like to process past submissions for AI writing, you will need to re-submit the document." Turning the feature on does not generate scores for work already handed in.
Does the Turnitin AI detection tool check paraphrasing tools?
Yes, and it is built into the same indicator. Turnitin describes detecting AI content "modified using a word spinner/AI paraphrasing or bypassing tool to evade detection", running automatically wherever AI writing detection is enabled. That layer is English only, unlike the main detector.
Does Turnitin check for AI in every language?
No. AI writing detection covers long-form English, Spanish and Japanese. Everything else returns no score. The paraphrasing and bypassing layer is narrower still: Turnitin says those capabilities "are only available for English submissions".
Can I see my own Turnitin AI score before submitting?
No. There is no student-facing AI score anywhere in Turnitin, and Draft Coach runs similarity, citation and grammar checks with no AI detection in them. The third-party sites offering a preview are separate companies running their own models, which cannot predict a Turnitin number.
Keep reading
The rest of our Turnitin coverage
Turnitin AI checker
Why no public Turnitin AI checker exists, and what the sites offering one actually are.
How accurate is the Turnitin AI detector?
Every accuracy figure traced to its source, including the ones with no traceable source at all.
How does Turnitin detect AI writing?
The segment-and-pool mechanism behind the percentage, step by step from Turnitin's own docs.
Does Turnitin detect QuillBot?
What the bypass and paraphrase layer is trained on, and why rewording often scores worse.
What AI percentage is allowed in Turnitin?
Why there is no universal pass mark, and what the asterisk in the 1 to 19 percent band means.
Best AI humanizer for Turnitin
What rewriting can realistically change before a Turnitin submission, and what it cannot touch.
Write drafts that read like a person wrote them
You cannot find out whether your school has AI detection switched on, so the only thing worth controlling is the writing. Our humanizer takes the flat, uniform register out of an AI draft so it reads the way you do.
Plans planned from $9/month billed yearly · Your runs saved to your history, delete any time