SafeAssign vs Turnitin: what Blackboard's built-in checker actually catches

If your institution runs Blackboard Learn, SafeAssign is probably the first line of defence you have against copied work — it's built in, it's free at the point of use, and most teachers assume it does roughly the same job as Turnitin. It doesn't, quite. The two tools solve overlapping but different problems, and the gap between them is worth understanding before you rely on either for a difficult case.

What SafeAssign actually checks

SafeAssign compares submitted text against four sources: the open internet, a ProQuest database of academic periodicals, Blackboard's own Global Reference Database of previously submitted student papers across institutions, and your institution's own archive of past submissions. When it finds overlapping strings of text, it reports a similarity percentage and highlights the matching passages.

That's a genuinely useful check for direct copying — a paragraph lifted from a website, an essay recycled from a previous semester, a paper shared between students at different schools that both use Blackboard. Where it struggles is anything outside those four sources: a paper bought from an essay mill, work translated from a non-English source, or content that's been paraphrased enough to break the string match. Similarity checkers in general — SafeAssign included — measure textual overlap, not originality of thought, so a well-cited literature review can score similarly to lightly-disguised copying.

Where SafeAssign and Turnitin genuinely differ

The biggest practical gap is database size. Turnitin's repository includes a much larger set of licensed journal and publisher content, plus a correspondingly larger base of institutions feeding it previously submitted student work. SafeAssign's ProQuest coverage and Global Reference Database are real, but narrower — which means the same plagiarised passage is measurably more likely to go undetected in SafeAssign than in Turnitin, simply because the source it was copied from isn't in the comparison set.

The second gap is AI detection. Turnitin added a dedicated AI-writing indicator to its similarity report in 2023. SafeAssign was not built around that feature in the same way — its core function remains textual similarity matching, not scoring how statistically 'AI-like' a passage of prose reads. If your institution is specifically trying to catch AI-generated text rather than copied text, SafeAssign on its own isn't the tool for that job, whatever your admin dashboard's default settings suggest.

What neither tool catches

It's worth being honest that this isn't really a SafeAssign problem — it's a similarity-checking problem, and Turnitin shares it. Neither tool can tell you whether a student who submitted an unflagged, 0%-similarity paper actually wrote it themselves. AI-generated text that draws on no single copyrighted source, paraphrased AI output run through a rewriting tool, or work written entirely by someone else for the student, all pass a similarity check cleanly. A clean SafeAssign report is evidence against direct copying — it is not evidence of authorship.

Adding a process signal alongside it

The evidence that actually answers 'did this student write this' isn't in the finished text at all — it's in how the work was produced. Genuine drafting has a shape: typing in bursts, pauses to think, some deletion and revision, a session length that roughly matches the difficulty of the task. Work that's pasted in wholesale, completed in a fraction of the expected time, or typed at an oddly even, unbroken rhythm tells a different story, regardless of what a similarity score says.

That's the gap Learnaway is built to fill alongside a similarity checker like SafeAssign rather than instead of one. It doesn't scan for matching text or score how 'AI-like' the prose reads — it records the timing and structure of the writing session itself, without ever capturing the words typed. A teacher can see that 90% of an essay arrived in a single paste two minutes into a six-minute session, which is a different — and often more useful — kind of evidence than a similarity percentage, especially for the submissions SafeAssign passes clean.

The practical takeaway

Keep SafeAssign running for what it's good at: catching direct copying from sources it can see. Don't treat a clean report as proof a student did their own thinking, and don't expect it to flag AI-generated text the way a dedicated AI checker would. If you're seeing submissions that pass every similarity check but still don't sit right, the writing process — not the finished text — is where the missing evidence usually is.