Yes. Packback runs its own AI Writing Detection feature natively, built to flag output from ChatGPT, Claude, Gemini, and other commercially available AI models. That puts it in a different category from most learning platforms covered on this site, which have no detector of their own and only check for AI writing if a school separately connects Turnitin or a similar tool.
What Packback's detector actually checks
The feature runs on both of Packback's products: Questions, its discussion-board tool, and Deep Dives, its longer-form essay tool. Packback says its detection method looks at repetitive phrasing and overly predictable sentence structure, a lack of original perspective or unique ideas, and writing-style inconsistencies measured against a student's own past submissions on the platform, not just against AI writing in general.
That last point needs stating carefully, because it is widely misdescribed. Packback's help centre does not document a per-student style profile, and an earlier version of this post said it did. What its AI Risk indicator actually examines is word choice and sentence structure, on the basis that AI models follow predictable patterns. Separately, its Uniqueness indicator matches a submission against Packback's internal corpus of previously submitted assignments, which does include your own earlier posts. That is a corpus match rather than a personal writing baseline, and the difference matters: the first can flag you for resembling other people's writing, the second for resembling your own.
How accurate is it, really?
Packback states its own false-positive rate at under 0.01%, but that's a vendor claim, not an independently audited figure, and no detector on the market, from any company, has a published number worth treating as gospel. The same caution applies here as with any AI detector: use the result as one signal, not a verdict, and don't assume a flag is automatically correct just because the platform advertises high accuracy.
The general reasons any AI detector, Packback's included, can misfire in both directions are covered in are AI detectors accurate.
How this compares to other platforms
Most learning management systems, Canvas, Brightspace, Google Classroom, have no AI checker of their own and only scan submissions if an instructor separately turns on a Turnitin integration.
See how that instructor-opt-in pattern works in does Canvas use Turnitin.
The same logic applies to does Google Classroom detect AI.
Packback breaks that pattern. Its detection runs as a built-in part of the product itself, which means it's on by default for any course using Packback rather than something an instructor has to separately enable.
- If your course uses Packback, assume AI detection is active on every submission unless your instructor has explicitly said otherwise.
- A clean pass on a different detector doesn't guarantee a clean pass on Packback's, since it's scoring against your own writing history, not just generic AI patterns.
Check your own draft with a free AI detector before you post it, as a second data point.
If it comes back flagged and the ideas are genuinely yours, our free AI Humanizer restructures the phrasing without changing what you're actually saying.
Frequently asked questions
Can I turn off Packback's AI detection for my own posts?
No. It's a built-in platform feature, not something an individual student can disable. Whether it's active for a given course is a platform-level setting, not a per-student one.
Does Packback only check discussion posts, or essays too?
Both. The AI Writing Detection feature runs on Questions, Packback's discussion tool, and on Deep Dives, its longer-form essay tool.
Is Packback's detector the same technology as Turnitin's?
No. They're separate products built by separate companies, using different detection methods. A result from one doesn't predict what the other would say about the same text.
What is the Originality Fingerprint, and what are its two numbers?
Packback packages its detection as the Originality Fingerprint, and understanding it means separating two numbers students routinely merge. Packback's help documentation, dated 28 October 2024, says the Originality Fingerprint appears in Discussion and Writing assignments, specifically Deep Dives and Originality Reviews, and is available for educators to view after a student submits. It contains two indicators, Uniqueness and AI Risk.
Uniqueness is the matching number: the percentage of your content that is not matched to other sources, checked against external sources found anywhere on the web and internal sources, which Packback defines as content from previously submitted Packback assignments. Two published caveats about that scan are worth more than any accuracy claim. Packback states the scan does not currently exclude quoted content or in text citations, so a properly quoted passage still drags your score down, and it states the scan does not currently include paywalled journal article sources, so a whole category of academic source is invisible to it.
Does Packback compare your writing to your own past work?
Partly, and not in the way it is usually described, including in the earlier version of this article. The distinction changes what you should worry about.
What Packback documents is that the Uniqueness check runs against internal sources, meaning previously submitted Packback assignments. Its FAQ describes that corpus as more than 50 million student submissions in Packback alongside over 100 million open web documents. Your own earlier Packback posts sit inside it, so recycling a paragraph you wrote yourself is a documented way to lower your Uniqueness score with no AI involved.
What Packback does not document is a per student writing style baseline feeding the AI Risk number. Its December 2024 help article says the algorithm analyses writing style by examining word choice and sentence structure, and that AI models tend to follow predictable patterns while human writing is more nuanced and varied. Nothing there says your submission is scored against a profile of how you personally write. Treat the claim that Packback knows your voice as unverified, and the claim that it has indexed everything you previously posted on Packback as documented.
What do Low, Moderate and High mean on the AI Risk indicator?
Packback's definitions are more hedged than the labels sound. Low means the algorithm did not detect any obvious AI generated text, with the caveat that a student may still have used AI and it is simply less likely. Moderate means it detected text that could have been generated by AI, and Packback's advice is to look closer and talk to the student. High means strong evidence of AI generated text, and Packback still says this does not necessarily mean the student intentionally used AI.
The most useful sentence Packback publishes is about who gets flagged wrongly: it says human writing can be flagged as moderate or even high risk, and that this is more likely with writing that is very concise, formulaic, or uses repetitive language. If you write short, tight, structurally repetitive discussion posts because that is what the prompt rewards, you are in the described false positive profile.
How seriously should you take Packback's false positive number?
Sceptically, and not because Packback is worse than its rivals. Packback publishes two different figures on its own site. Its platform page claims the lowest false positive rate of any commercially available tool at under 0.01 percent. Its FAQ says the model was tuned to a 0.005 percent false positive rate, translated as 1 out of every 20,000 submissions. Both are vendor claims, neither is independently audited, and they are not the same number. Checked August 2026.
The model behind them has changed too. Packback published a note dated 30 January 2025 saying it had transitioned to a new AI writing detection model and that earlier reports can be refreshed by clicking Regenerate, so a report generated last term and one regenerated today are not guaranteed to agree.
What else do students ask about Packback's detector?
Can I see my Originality Fingerprint before I submit?
Sometimes. Packback's student documentation says that depending on the settings for your assignment you may be able to generate or view your Originality Fingerprint before submitting, and that if you do not have access you should ask your educator, who may be able to turn it on.
Will quoting and citing properly still lower my Uniqueness score?
Yes. Packback states its scan does not currently exclude quoted content or in text citations, so a correctly attributed block quotation still counts as matched text. A low Uniqueness score is not evidence of anything on its own; the report showing the matched sources is what matters.
