
Short answer: no. SafeAssign does not detect AI writing. It matches your text against existing sources and gives no AI score. Its maker, Blackboard (called Anthology until March 2026), tested AI detection for SafeAssign in 2023 with 65 client partners and judged it "not currently fit for purpose in education." AI scores in Blackboard come from added tools such as Turnitin.
I read 6 Blackboard help pages, its March 2026 release notes and its AI white paper. SafeAssign does not detect AI writing on Blackboard's own account.
Does SafeAssign detect AI or only matching text?
Only matching text. Blackboard says SafeAssign uses "a unique text matching algorithm capable of detecting exact and inexact matching between a paper and source material." It checks 4 sources, and the largest, the Global Reference Database, holds over 15 million papers volunteered by students. Google Classroom's originality reports work the same way, as our Google Classroom AI guide explains.
The others are your institution's past submissions, the internet and ProQuest's ABI/Inform journals (Blackboard). SafeAssign does not detect AI writing; this table shows every other layer in a Blackboard course.
| Layer | What it looks at | AI score? |
|---|---|---|
| SafeAssign Originality Report | Text matching existing sources | No |
| Document Properties (Ultra, March 2026) | File author, dates, revisions, software | No |
| AI Design Assistant | Course content for instructors | No |
| Turnitin, if connected | Similarity and AI writing | Yes, where enabled |
| Cursive, if added | Typing, edits, time on task | No, a process record |
| Your instructor | Your writing across the term | Human judgement |
Why did Blackboard turn down AI detection for SafeAssign?
Its own test went badly. In May and June 2023, 65 client partners submitted more than 1,000 human and AI texts to detectors tested for SafeAssign. Afterwards, 80% of respondents felt the detectors could at best "sometimes" identify texts correctly, leaving 20% who said often or almost always.
The white paper also records bias. One respondent wrote: "ESL and ASD student work more likely to be flagged than actual AI." Blackboard said it did not expect AI detection "will reach the required level of accuracy for ethical use any time soon."
Its Protecting Against AI Plagiarism page, updated 30 September 2026, still links that paper, so SafeAssign does not detect AI writing. Our roundup of whether AI detectors are accurate covers wider studies.
Does SafeAssign detect AI-paraphrased text?
Not as AI. SafeAssign catches "exact and inexact matching", so text reworded from a source in its 4 databases can still show up as a match. Text a chatbot writes from scratch usually has no source to match, a gap Blackboard's 2023 white paper names in 1 sentence.
Generative AI "creates original text and media that appear authentic and do not match with any existing work," it says. A low score clears nobody. If the worry is reworded text, Turnitin is the tool that tries to spot it, as our guide on how AI detectors work explains.
Where else can Blackboard surface AI writing?
Not natively. Since March 2026, SafeAssign reports in Ultra courses include a Document Properties panel with up to 7 metadata fields, such as file creation date, revision count and software used. Blackboard's release notes say these "complement existing detection methods". They aren't an AI score.
Metadata describes the file, not the writing process.
Blackboard's Cursive records "typing behavior, edit history, and time on task" through a browser extension (Blackboard). Matthew Darch of Murdoch University says academics reported it gave them "far greater confidence that students' writing was genuinely their own."
Where does an AI score in Blackboard come from?
From a connected tool such as Turnitin. Turnitin's FAQ says "Blackboard has already fully transitioned customers to LTI 1.3", and an AI Writing column appears only where the institution has enabled AI writing detection. Turnitin needs at least 300 words of prose.
Instructors must accept a disclaimer before they see AI scores (Turnitin). Its AI Writing Report guide says the AI score "should not be used as the sole basis for adverse actions against a student." Canvas works the same way (can Canvas detect AI), and our Turnitin false positive guide covers false flags. On Moodle instead of Blackboard? Our Moodle AI detection guide covers its plugins.
Does SafeAssign detect AI in tests, forums and PDFs?
SafeAssign does not detect AI writing in tests, forums or PDFs. Tests get a report that includes text questions. Blackboard's FAQ names 3 places SafeAssign doesn't run: forums, peer assessment and Content Collections. It also won't compare a new attempt with your earlier ones or check work sent before it was switched on.
It "does not currently perform OCR", so scanned PDFs can't be read (Blackboard), and its internet index can lag by 1 to 3 days (Blackboard). That's no loophole: a missing report stands out, and Blackboard tells instructors to "observe each learner's writing style."
What should you do with a high SafeAssign score?
Read the matches first. Blackboard calls the overall score "a warning indicator only" and sorts papers into 3 bands: below 15% is low, 15% to 40% is medium and over 40% is high. Papers with extensive quoted or paraphrased material sit in the medium band, says its Originality Report guide.
Then run 4 checks:
- Ask if your instructor shares SafeAssign results.
- Open the report "shortly after you submit your attempt."
- Check each highlighted block is quoted and cited.
- Keep your drafts.
If you're questioned about AI rather than copying, see how to prove you didn't use AI.
FAQ
Does SafeAssign detect AI writing?
No. It only reports text that matches its sources.
Does Blackboard detect AI without Turnitin?
No. Any AI percentage comes from a tool your school connects.
Can students see their SafeAssign report?
Only if the instructor allows it for that assessment.
What do Ultra Document Properties show?
File details such as revision count. Ultra courses show them as context only, not as an AI score.
Are discussion posts checked by SafeAssign?
No. SafeAssign doesn't run in forums at all.


