What does SafeAssign check for?
Matching text. Anthology’s student overview says SafeAssign “is based on a unique text matching algorithm capable of detecting exact and inexact matching between a paper and source material,” and that assignments “are compared against several different databases containing millions of articles dating from the 1990s to the present.” It checks every part of what you hand in: SafeAssign “creates an Originality Report for each part of your assignment or test, including text questions and attachments.”
Your instructor turns it on for an assignment or test, and you don’t do anything extra. Anthology says assignments using SafeAssign “don’t require you to take any extra steps when you’re ready to submit your work.” You can tell it is on if you “see Originality Report enabled in the Details & Information” panel.
| Compared with | What Anthology says |
|---|---|
| Published sources and the web | Its database holds “academic papers, publications, articles, and internet sources,” and it “searches its database and the internet” for matches (white paper). |
| Your school’s earlier submissions | Submissions are added to the Institutional database unless your instructor chose to exclude them before you submitted. |
| Other schools’ papers | The Global Reference Database “is a separate database from your institution’s database,” and “papers from other institutions are checked against your paper.” |
| Deleted submissions | “When a student’s submission is deleted from a course, the content remains in the database and can be used to detect future matches for other student submissions.” |
| Your own earlier attempts | Not matched. SafeAssign “doesn’t check the contents of the current attempt against content from previous submissions.” |
Does SafeAssign detect ChatGPT or other AI writing?
No AI check appears in the SafeAssign help we read on October 20, 2026. The submit guide, the student overview, the Originality Report guide and the instructor FAQ all describe matching text against sources, and the report’s numbers are match percentages. None of them describes an AI-likelihood score.
Anthology explains the gap in a white paper that its instructor help still links, AI, Academic Integrity, and Authentic Assessment. It says common anti-plagiarism tools “were built to compare learners’ work with existing texts,” while generative AI “creates original text and media that appear authentic and do not match with any existing work.”
Anthology also tested adding AI detection. “As part of an in-depth process to evaluate the inclusion of AI detection within SafeAssign,” it ran a beta with “market leading AI detection solutions” in May and June 2023, in which 65 of its client partners submitted more than 1,000 texts. Its conclusion was that “AI detection is not currently fit for purpose in education.” The paper adds that Anthology does not expect AI detection “will reach the required level of accuracy for ethical use any time soon,” and that it will “continue to monitor this closely.”
The paper dates from 2023. Anthology’s help page Protecting Against AI Plagiarism, last modified September 30, 2026, still points to it and says it covers “using anti-plagiarism tools such as SafeAssign versus AI detectors.” Features can change, so if your syllabus names a tool, believe the syllabus.
How do you read a SafeAssign Originality Report?
It shows how much of your text matches other sources, and where. Anthology’s student guide lists three parts in the preview:
- Overall Risk shows “whether your attempt is a low, medium, or high risk for plagiarism,” based on your average match.
- Overall Text Matching shows “the average and highest text match for the pieces of your attempt.”
- Originality Reports shows “the percent of potential matching text for each piece of your submission, including questions and attachments.”
Anthology gives these score bands for the overall score:
| Overall score | Risk | What Anthology says these papers include |
|---|---|---|
| Below 15% | Low | “some quotes and few common phrases or blocks of text that match other documents” |
| 15% to 40% | Medium | “extensive quoted or paraphrased material, or they may include unoriginal content” |
| Over 40% | High | “A very high probability exists that text in these papers was copied from other sources” |
A high number is a prompt to look closer. Anthology calls the overall score “a warning indicator only,” and its student overview says the report “does not state whether a phrase that matches a source is properly referenced. Your instructor must read the report and determine if you used proper citations.”
What else can your instructor see besides the match score?
More than a percentage. Three things on Blackboard’s own pages:
- Document properties. The full report has a Document Properties section with file metadata “such as” the author or creator, the file’s creation date and time, who last modified it and when, the document revision count and the software used. Anthology says it respects anonymous grading, and fields that aren’t available aren’t shown.
- Your writing. Anthology’s advice to instructors is to set personal, reflective tasks, which let them “observe each learner’s writing style, making it easier to spot any AI-generated work that the student may submit later.”
- Other tools your school adds. Schools on Blackboard have access to SafeAssign “as well as native integrations with tools from major antiplagiarism providers,” the white paper says. Blackboard also sells Cursive, a separate authorship product whose page describes “capturing typing behavior, edit history, and time on task.” That page says Cursive connects through a browser extension today. Whether any of these runs in your course is up to your school.
Can you see your own SafeAssign report?
Only if your instructor allows it. “Your instructor decides whether you can see Originality Reports for your assignment attempt.” If you can, Anthology says “you can check if you properly cited your sources,” and you can open the report before grading: in the assignment’s Details & Information panel, select View Originality Report. With more than one attempt allowed, “SafeAssign analyzes all of your attempts separately,” and a new attempt is not matched against your earlier ones.
One choice is your instructor’s alone. To keep submissions out of the Institutional and Global Reference Databases, Anthology’s FAQ says the instructor must select the exclude option “before students submit their work.” Work submitted earlier can’t be removed from them.
Is using AI on a SafeAssign assignment allowed?
SafeAssign doesn’t decide that. Your instructor and school do, and Anthology’s white paper argues that instructors “should also have autonomy” to apply AI guidelines to their subject. The rule for your course is in the syllabus or your school’s academic-integrity policy, and it applies whether or not any tool flags anything. If it isn’t clear, ask your instructor before you submit. We wrote more on that question in Is using an AI homework bot cheating?
Other platforms handle this their own way. Packback, for example, shows an AI Risk indicator next to your own discussion posts, which Does Packback detect AI? explains from Packback’s help center. For the rest of Blackboard, including what its test logs record, see Silent Student on Blackboard.
Where Silent Student fits
We make Silent Student, a Mac and Windows desktop app from El Dorado Ignition LLC with a Blackboard Learn Ultra mode. It submits through Blackboard’s own pages, so an assignment with SafeAssign turned on is checked like any other submission. It does not hide work from SafeAssign or any other tool, and your school’s academic-integrity policy applies.
More: Silent Student on Blackboard.
Questions people ask
Does SafeAssign detect AI?
Anthology’s SafeAssign help describes matching your text against published sources and other papers, not AI detection. Its 2023 white paper says it tested AI detection for SafeAssign and found it not fit for purpose in education.
Does SafeAssign detect ChatGPT?
Not as a separate check in the help we read. SafeAssign reports text that matches its sources. Your instructor still reads your writing and the report’s document properties, and your school may use other tools.
What is a bad SafeAssign score?
Anthology calls an overall score below 15% low risk, 15% to 40% medium and over 40% high. It also calls the score a warning indicator only: your instructor decides whether matches are properly cited.
Does SafeAssign check my work against my earlier attempts?
No. Anthology says SafeAssign treats a new attempt as coming from the same student and doesn’t check it against your previous attempts.
Does SafeAssign keep my paper?
Usually. Unless your instructor excluded submissions from the databases before you submitted, your paper stays in SafeAssign’s database and is used to find matches in later work, even if it is deleted from the course.
Can I see my SafeAssign score before it is graded?
If your instructor allows it. Anthology says you can open the Originality Report before grading from the assignment’s Details & Information panel, but the instructor decides whether students see it.
Sources
Checked October 2026: each page below was fetched and read on October 20, 2026.
- Submit with SafeAssign (Anthology, student help, last modified October 6, 2026)
- SafeAssign (Anthology, student help)
- SafeAssign Originality Report (Anthology, student help, last modified September 30, 2026)
- SafeAssign FAQs (Anthology, instructor help)
- Protecting Against AI Plagiarism (Anthology, instructor help, last modified September 30, 2026)
- AI, Academic Integrity, and Authentic Assessment: An Ethical Path Forward for Education (Anthology white paper, landing page)
- AI, Academic Integrity, and Authentic Assessment (PDF) (Anthology white paper, beta test of May and June 2023)
- Cursive by Blackboard (Blackboard product page)
How this guide was made. We read each Anthology and Blackboard page listed above on October 20, 2026, including the full white paper PDF, and quote them directly. None of the SafeAssign help pages listed above describes an AI-detection feature. We wrote the guide with AI assistance and checked each quote against its source.
Related guides: What can teachers see on Canvas? · Silent Student on Blackboard · Does Packback detect AI? · Is using an AI homework bot cheating? · Blackboard grade calculator · All guides
Written by the company that makes Silent Student. Silent Student is not affiliated with or endorsed by Anthology or Blackboard. Blackboard, SafeAssign and Cursive are trademarks of their respective owners.