QuillBot is the tool students ask about by name, so it deserves an answer by name. The short version: Turnitin has specifically updated its detector to identify AI-paraphrased text — which is precisely what QuillBot produces — and the workflows people use QuillBot for fail in two different ways depending on what's being paraphrased. Let's separate them, because the risk profiles are different.
Workflow 1: paraphrasing AI-generated text
This is the "ChatGPT writes it, QuillBot launders it" pipeline, and it's the one Turnitin's AI-paraphrasing detection was built for. Paraphrasers change vocabulary and shuffle clauses, but the underlying machine-shaped structure — even rhythm, predictable progression — largely survives, and detectors have been trained on exactly this before-and-after pattern. Turnitin's detector can flag such text both as AI-generated and, in its newer versions, as AI-generated-then-paraphrased. The paraphrasing-tool arms race is real, but it isn't one you're positioned to win the week your grade depends on it.
Workflow 2: paraphrasing sources to dodge similarity
The older QuillBot use case — running a source paragraph through the paraphraser so it won't match — has a different failure mode. Synonym-swapped text with the source's sentence structure intact is patchwriting, the exact pattern markers are trained to spot when they read the match list, and partial matches often survive anyway. Worse, you've now added machine-generated phrasing on top, so text that started as uncited copying can pick up an AI flag too. One shortcut, two detectors.
What about the "Humanizer" mode?
QuillBot and similar tools now market modes that promise to make AI text undetectable. Treat every such claim as a screenshot from a specific week: detectors update, and a bypass demonstrated against last semester's model says nothing about the version your university runs today. Nobody selling a humanizer will refund your misconduct hearing. If the text matters, the only evidence worth anything is the actual Turnitin report on your actual document.
The legitimate uses, for completeness
Fairness requires saying: QuillBot also gets used in ways nobody objects to — rephrasing your own clunky sentence to see alternatives, tightening wordy phrasing, English learners studying how an idea can be restructured. The line, as with Grammarly, is authorship: a tool suggesting a better version of a sentence you wrote is editing; a tool producing your paragraphs from someone else's (or a model's) text is not. And your course's AI policy, not any detector, defines what's allowed.
The bottom line
Does Turnitin detect QuillBot? It detects what QuillBot leaves behind: AI-paraphrased text on the AI report, and structure-preserved paraphrase on the similarity side. If you've used it on a document that matters and you're now unsure where you stand, stop guessing — run the real check in no-repository mode, read both reports, and fix what's actually flagged rather than what forums think might be.