What AI detector do colleges use?

Platform by platform, as documented on 17 August 2026

The learning platform and the detector are two different purchases. Every claim below comes from the vendor's own documentation, checked on the date this page was published — these products change often enough that the date matters as much as the answer.

• Canvas (Instructure) — no first-party AI-writing detection. Canvas provides a plagiarism framework that third-party tools plug into, so what appears in a Canvas assignment is whichever tool the institution connected: Turnitin most often, with GPTZero and Copyleaks also offering integrations.
• Blackboard SafeAssign (Anthology) — text matching only. Anthology's own documentation says SafeAssign "compares your submitted assignments against a set of academic papers to identify areas of overlap" across databases dating from the 1990s. It makes no AI-detection claim at all.
• Google Classroom originality reports — plagiarism matching, not AI detection. Google describes the comparison set as "hundreds of billions of web pages and over 40 million books", plus an optional school-owned repository of past student work on Education Plus and the Teaching and Learning Upgrade. AI-generated writing is not mentioned.
• Turnitin — the actual detector in most cases. Its AI writing indicator lives inside the Similarity Report and is available only to institutions licensing Turnitin Originality. Administrators can disable it, and if they do, nothing is processed for AI writing.
• College Board — AP Capstone and AP Computer Science Principles submissions go through Turnitin originality reports in the AP Digital Portfolio, and flags cover plagiarism, fabrication and unacceptable use of AI.
• College admissions — no detector is named by anyone. The Common App fraud policy defines fraud to include misrepresenting as your own work "the substantive content or output of an artificial intelligence platform, technology, or algorithm", but investigations are triggered by reports, and each member college applies its own internal protocols.

Why "does Canvas have an AI detector" gets contradictory answers

Because the honest answer is both. Canvas has no detector; a Canvas assignment can still return an AI score. The score is generated by Turnitin and read inside Turnitin's report, which the student reaches through Canvas — so it looks like a Canvas feature and is not one.

Turnitin states this directly: users see the indicator and the report via the LMS they are using, but "There is no AI writing indicator or score embedded directly in the LMS user interface" — you have to open the report to see the score. The same applies to Blackboard and Moodle. It also means the answer to "does my college use an AI detector" cannot be settled by looking at the LMS. Two courses on the same Canvas instance can be configured differently.

What Turnitin's AI indicator actually shows, and who can see it

A percentage of the qualifying text that the model predicts was AI-generated or AI-paraphrased, visible to instructors and administrators only. Students do not see it, though an instructor can download the report as a PDF and share it.

Three details from Turnitin's FAQ that change how the number should be read. Scores between 1% and 19% are shown as an asterisk with no percentage attributed, deliberately, to avoid false positives at the low end. The model only analyses long-form prose — not bullet points, not lists, not code — so the percentage is a share of the qualifying text rather than of the whole document, which is why the highlighted text often looks like less than the number suggests. And a submission needs at least 300 words of prose to be processed at all, under 30,000 words, under 100 MB, as .docx, .pdf, .txt or .rtf.

The accuracy claim, stated as Turnitin states it

Turnitin claims a false positive rate under 1% for documents containing over 20% AI writing, and says it validates that figure before each model release by testing on more than 700,000 academic papers written before ChatGPT was released.

Two things sit alongside that claim. Turnitin accepts it misses AI writing as the price of the low false-positive rate — its own example is that a document scored at 50% could contain as much as 65% AI writing. And it says plainly that it does not determine misconduct: "the percentage on the AI writing indicator should not be used as the sole basis for action or a definitive grading measure by instructors." Under 1% is also not zero. Across a large course, some human-written papers will be flagged, which is the whole reason the appeal question below matters.

Does using Grammarly get you flagged?

Turnitin says its detector is not tuned to flag spelling, grammar and punctuation corrections, and that in testing, changes made by Grammarly free and premium were in most cases not flagged. Its generative features are a different matter.

Turnitin explicitly excludes "content generated by Grammarly's generative AI-powered features, including draft generation, paraphrasing, summarizing" from that reassurance, and says such content will likely be flagged as AI. The practical line is between a tool correcting sentences you wrote and a tool writing sentences for you — which is the same line most institutional policies draw, and the reason to read your own before relying on either.

How to find out what your own institution runs

Ask, or read the policy. The academic integrity section of your syllabus or student handbook is the authoritative answer for you, and it is the only one that carries any weight in a dispute.

Three places worth checking, in order: the course syllabus and the institution's academic integrity policy, which increasingly name the tool and say whether AI use is permitted, restricted or banned; the assignment settings in your LMS, where an enabled plagiarism tool is usually visible when you submit; and your instructor, who can simply tell you. What is not worth checking is a forum post about what some college used last year. Licensing changes, administrators toggle the feature off, and departments set their own rules.

If your work has been flagged and you did write it

Keep the evidence of how it was written, and use the appeal route the institution provides. Document history is the strongest thing you can produce, and it only exists if you did not paste the whole draft in at once.

Version history in Google Docs or Word, your notes and outlines, browser or library records of the sources you read, and earlier drafts all show a document being built rather than arriving. The formal routes differ: College Board tells students flagged in the AP Digital Portfolio to reply to the notification email and include supporting evidence, while most colleges run an academic integrity process with a stated appeal step. Turnitin's own position — that the score is data for an educator to weigh, not a determination of misconduct — is worth quoting in that conversation, because it is the vendor's position and not the student's opinion.

Where our app fits, and where it does not

AI Humanizer runs a detection score on a draft and shows which sentences read as mechanical, then helps you rewrite them. It is a writing tool for people editing their own work — three free actions with a 300-word limit before paid plans, on iPhone.

What it is not: a way to submit AI-written work as your own. Our own FAQ refuses to claim your text will pass a detector, because detectors are inconsistent and change constantly — the model list Turnitin publishes gets longer every few months. If your institution bans AI-assisted writing, running a draft through any tool of this kind, ours included, may itself breach the policy. Read the policy first; that decision is not one an app can make for you.