SafeAssign vs Turnitin AI Detection (2026 Comparison)

SafeAssign doesn't detect AI writing. It's Blackboard's text-matching tool: it compares a paper with its reference databases and the internet and reports how much of it overlaps with existing sources (Blackboard help). Turnitin does that kind of matching too, and schools that license Turnitin Originality also get a separate AI writing indicator, which Turnitin activated on 4 April 2023. So the honest answer to "which one catches more AI?" is that only one of them tries.

That's the key thing to know if your course runs on Blackboard: a SafeAssign Originality Report is a match percentage, not an AI score. Below is how each tool works according to its own documentation (checked September 2026), what each score means, what's known about false positives, and how to tell which one your course uses.

Quick Comparison Table

Sources: SafeAssign (instructor help, sources, student help); Turnitin (AI writing FAQ, Similarity Report guide, LMS guides).

The comparison people usually want, a "detection rate" for each tool, doesn't exist for SafeAssign, because SafeAssign doesn't score AI. And no independent study we could find has measured Turnitin's current AI model on a public test set, because Turnitin only sells to institutions.

How SafeAssign Works

Blackboard describes SafeAssign as a tool that "compares submitted assignments against a set of academic papers to identify areas of overlap between the submitted assignment and existing works" (SafeAssign instructor help). According to the SafeAssign sources page, it searches three pools of text:

• the Global Reference Database, which holds "over 15 million papers volunteered by students from Blackboard client institutions";
• your institution's archive of papers previously submitted to SafeAssign;
• the internet, through an internal search service.

Its algorithm looks for "exact and inexact matching" between your paper and those sources. That's similarity checking. A low score means the text doesn't overlap much with those sources. It says nothing about whether a person or a chatbot wrote it, and Blackboard's SafeAssign help pages don't mention AI-generated writing at all.

The Originality Report guide sorts scores into three bands. Below 15 percent is low ("some quotes and few common phrases"), 15 to 40 percent is medium ("extensive quoted or paraphrased material"), and over 40 percent is high. The same page calls the overall score "a warning indicator only" and tells instructors to "Review papers to see if the matches are properly attributed."

SafeAssign only runs where the instructor turns it on. In Ultra courses, the instructor enables the Originality Report in the assignment settings, and can separately choose to "Allow students to view the Originality Report" (SafeAssign in Your Course).

The 2026 change: document metadata, not AI

In March 2026, Blackboard added a Document Properties section to the Originality Report. According to the March 2026 update post, it shows "the document author, creation and modification dates, and the document revision count." That's information about your file, not an analysis of your prose, and a low revision count proves nothing on its own. It's still worth checking your file's properties before you upload it.

Why SafeAssign Has No AI Detector

On 18 December 2023, a post on blackboard.com announced that "Anthology has decided not to make AI plagiarism detection available natively within Blackboard Learn." The reason it gave: "Our extensive research has shown that AI detection models are biased and unreliable, and thus inappropriate to use in such a high-stakes workflow."

That decision followed a trial. On 1 June 2023, GPTZero announced a beta that let instructors at "a few dozen of Anthology's US-based Blackboard Learn clients" use GPTZero inside SafeAssign. GPTZero's February 2026 LMS comparison still says its detection "works from within SafeAssign in Blackboard LMS," but Blackboard's own SafeAssign documentation doesn't describe any AI component. If someone tells you SafeAssign checks for AI, ask which product actually produces that score. We cover the Blackboard side in more detail in does Blackboard detect AI?

How Turnitin's AI Detection Works

Everything in this section comes from Turnitin's AI writing detection FAQ, checked in September 2026, unless another link is given.

• It scores segments, not the whole file at once. Sentences are "segmented into overlapping sections," and each segment gets "a value between 0 and 1, denoting the probability of the text being likely human or AI-generated." Turnitin says its classifiers are "trained to detect these differences in word probability."
• It reports a share of the text. The model "calculates the overall percentage of text in the submitted document that was likely generated by an AI writing tool." That's a different kind of number from SafeAssign's match percentage.
• It needs 300 words of prose. A file "must have at least 300 words of prose text in a long-form writing format" and must not exceed 30,000 words.
• Low scores become an asterisk. "To avoid potential incidence of false positives, no score or highlights are attributed for AI detection scores in the 1% to 19% range."
• Four languages. Long-form English, Spanish, Japanese and Modern Arabic. Its "AI paraphrasing and bypassing detection capabilities are only available for English submissions."
• One model since July 2026. "In July 2026, we updated our model architecture to consolidate a multi-model ensemble into a single model." We walk through the report changes in Turnitin's August 2026 update.
• Named models. For English, Turnitin lists models it can detect, from GPT-4o and the GPT-5 family to Gemini, Claude, LLaMA, Mistral, DeepSeek and Grok versions, "and tools based on these LLMs as well."
• It's a licence add-on. "AI writing detection is only be available to customers that license Turnitin Originality." A Turnitin assignment in your Blackboard course doesn't automatically mean an AI check.
• Students don't see it. "The AI writing detection indicator and report are not visible to students," though instructors can download and share the report. More in can students see their Turnitin AI score?
• Not the sole basis for action. Turnitin says the percentage "should not be used as the sole basis for action or a definitive grading measure by instructors."

Two Different Numbers: Match Score vs AI Score

The most common mix-up is treating a SafeAssign percentage as an AI percentage. They measure different things:

Turnitin itself keeps the two separate: its Similarity Report and its AI indicator are different reports. Our guide to similarity score vs AI score walks through both.

False Positives: What's Documented

For SafeAssign, a "false positive" means a match that isn't plagiarism, such as a common phrase or a correctly cited quotation. That's why Blackboard tells instructors to review whether matches are "properly attributed" before drawing conclusions.

For Turnitin's AI score, Turnitin states its target: it keeps its false-positive rate "under 1% for documents with over 20% of AI writing," which it glosses as flagging "one out of every 100 fully-human written documents." It says it tests "over 700,000 additional academic papers that were written before the release of ChatGPT" before each update, and accepts missing some AI text to keep false positives low: a document scored at 50% "could contain as much as 65% AI writing" (FAQ). Those are Turnitin's own figures.

Independent checks are older and smaller:

• In a 2023 peer-reviewed test of 14 tools run in March and April 2023, Turnitin classified the most documents correctly (41 of 54) and made no false accusations on the 18 human-written or machine-translated documents.
• In a preprint by Perkins et al., testing in September and October 2023, Turnitin flagged none of the 10 human samples, but its accuracy fell by 42.1 percentage points when AI text had been altered to evade detection, the largest drop of the seven detectors.
• Vanderbilt did the scale arithmetic when it disabled Turnitin's AI detector on 16 August 2023: at the 1% rate Turnitin claimed at launch, and 75,000 papers submitted in 2022, "around 750 student papers could have been incorrectly labeled."

More cases and studies are in our explainer on AI detection false positives.

Which One Is Your Course Using?

Which tool you face depends on your institution's licences and your instructor's settings, not simply on the LMS. Turnitin publishes integration guides for Blackboard as well as Canvas, D2L Brightspace, Moodle, Schoology, Google Classroom and Microsoft Teams (Turnitin in an LMS), and Turnitin says over 16,000 academic institutions, publishers and corporations use its services (Turnitin). Some institutions license Turnitin but have switched its AI indicator off, including Vanderbilt (August 2023) and the University of Waterloo (from September 2025). We track others in universities dropping AI detection.

To find out what applies to your assignment:

1. Read the assignment details. When SafeAssign is on, Blackboard's student help says "Originality Report enabled" appears in the details. That's a similarity check.
2. Look for Turnitin's name. Turnitin assignments are set up through Turnitin's own tool, so they usually carry its name.
3. Ask directly. "Does this go through SafeAssign only, or Turnitin's AI report too?" is a fair question for an instructor.
4. Check the syllabus for an AI policy. What's allowed is set by your course, not by either tool.

Can Either One Be "Bypassed"?

Students search for ways to bypass SafeAssign and Turnitin constantly, so it's worth being precise about what each tool does.

SafeAssign has no AI score to lower. What it flags is text that matches existing sources. The honest way to keep that score meaningful is to quote and cite properly, because the report is built to show matches, and your instructor then decides whether they're attributed.

Turnitin says its AI detection includes "likely AI-generated content that may have been modified using a word spinner/AI paraphrasing or bypassing tool," run automatically on English submissions where the AI indicator is enabled (FAQ). It announced that capability on 27 August 2025. Independent research shows detectors can be fooled, but nobody outside Turnitin can test its current model, and no tool can promise how it will score a particular paper.

If you use AI tools in ways your course allows, keep your drafts and version history, and disclose the use as your policy requires. SupWriter's AI detector can give you a rough pre-check before you submit, but it's one scoring model, not SafeAssign or Turnitin, and it can be wrong (how we test).

Which Platform Is Better?

It depends on the job.

For catching copied or unattributed text: both do it, against different sources. SafeAssign draws on its student-paper database, your institution's archive and the internet. Turnitin adds a subscription repository of periodicals, journals and publications to the internet and its repository of previously submitted papers. Neither publishes a comparative match rate, so we can't say which finds more.

For AI writing: only Turnitin produces a score, and only on Originality licences, in four languages, for 300 or more words of prose. Blackboard's parent company chose not to build one, calling AI detection models "biased and unreliable."

For students: neither report is a verdict. A SafeAssign score needs a human to review the matches, and Turnitin says its AI percentage shouldn't be the sole basis for action. What protects you in either case is the same: drafts, version history, notes and sources. Our guide on proving your writing is human covers what to gather.

FAQ

Does SafeAssign detect ChatGPT?

No. Blackboard documents SafeAssign as exact and inexact text matching against its reference databases and the internet, and Anthology said in December 2023 that it had decided not to offer AI detection natively in Blackboard Learn (Blackboard).

Can my professor see an AI score in Blackboard?

Only if the course uses a separate tool, such as Turnitin with an Originality licence. Turnitin's AI percentage sits inside the Turnitin report, not in SafeAssign.

Is a high SafeAssign score an AI flag?

No. Blackboard's bands are about overlap: under 15% low, 15% to 40% medium, over 40% high, and the score is "a warning indicator only" (Originality Report guide).

What AI percentage gets flagged on Turnitin?

Turnitin doesn't set one. It shows 1% to 19% only as an asterisk and leaves decisions to institutions. See what percentage of AI is acceptable on Turnitin.

---

Want a rough read before you submit? Check a draft with SupWriter's AI detector. 300 words free, no credit card required.