Direct answer
Turnitin does detect text produced by chat models such as ChatGPT, Claude, and Gemini, and it reports that detection as an AI writing percentage plus sentence-level highlighting inside the same report your instructor already uses for similarity [1]. The feature is not a separate "chat detector" product with its own login; it is a layer inside the standard Turnitin feedback workflow, so a chat-written paragraph and a copied paragraph can surface in one document review. That matters because most students who search for a chat detector are not looking for a verdict on the tool — they are trying to understand what a marker will see on their own draft [1]. This guide explains what the detector looks for, how the report communicates its findings, and how a pre-submission check fits into that picture.
Introduction
Turnitin does detect text produced by chat models such as ChatGPT, Claude, and Gemini, and it reports that detection as an AI writing percentage plus sentence-level highlighting inside the same report your instructor already uses for similarity [1]. The feature is not a separate "chat detector" product with its own login; it is a layer inside the standard Turnitin feedback workflow, so a chat-written paragraph and a copied paragraph can surface in one document review. That matters because most students who search for a chat detector are not looking for a verdict on the tool — they are trying to understand what a marker will see on their own draft [1]. This guide explains what the detector looks for, how the report communicates its findings, and how a pre-submission check fits into that picture.
Does Turnitin detect text written by ChatGPT and other chat models?
Yes. Turnitin's AI writing detection is built to identify patterns typical of generative models, and it is applied to submissions written with ChatGPT, Claude, Gemini, and comparable chat assistants [2]. The output is not a binary "AI / not AI" stamp. It is a percentage covering the share of the document flagged as likely AI-generated, paired with highlighting so a reader can move from the number to the specific sentences behind it [2]. In practice this means two drafts can both score in the same band while one is flagged in a single paragraph and the other across an entire section — the number alone tells you very little without the highlights [2].
It is also worth separating what the detector reads from what it does not. The analysis works on the text itself, so it responds to phrasing, sentence rhythm, and structural regularity rather than to whether a chat window was ever opened [2]. A student who drafted in their own words but asked a model to smooth a paragraph can still see that paragraph highlighted, while a heavily edited chat draft may come back with far less flagged [2]. Because the report describes text and not intent, Turnitin frames its output as a signal for human judgement rather than a finding of misconduct [2].
One practical consequence is that results are not perfectly stable across submissions. The same passage resubmitted in a slightly different document can land on a different percentage, particularly near the boundary of the low-confidence band [2]. That variability is a reason to treat any single number as a snapshot of one draft on one day, not as a permanent label attached to a student [2].
How does the Turnitin AI writing report actually flag chat-generated sentences?
The report communicates its findings through two layers that are meant to be read together: an overall percentage for the submission and colour-coded highlighting on the sentences that drove that percentage [3]. Highlighted segments represent passages the system considers likely to have been generated by AI, and the report distinguishes between text judged AI-generated and text judged to show signs of AI paraphrase or light rewriting [3]. A reader who clicks into a highlight is looking at the exact sentence the model reacted to, which is why the highlight view is usually more informative than the headline score [3].
The percentage has a documented floor that surprises many students. When the detected share falls below Turnitin's confidence threshold of roughly 20%, the report displays an asterisk form — *% — instead of an exact figure, signalling a low-confidence result rather than a clean bill of health [3]. Seeing *% therefore does not mean "zero AI"; it means the signal was too weak to report as a precise number, and a reviewer may still open the highlights and read them [3]. Understanding that distinction prevents the common misreading of a low score as proof that nothing will be questioned [3].
How the report is used sits outside the algorithm entirely. Turnitin's own guidance directs instructors to treat highlights as a starting point for a conversation about the writing process rather than as evidence for disciplinary action [3]. Many institutions configure the report so that students can see it alongside feedback, while others keep it instructor-only, which is why two classmates in the same cohort can have very different experiences of the same detector [3]. The report is a reading aid; the interpretation is institutional.
Can I check my own draft for AI chat text before I submit it?
Not through Turnitin directly. The AI writing report is generated inside an institutional account, and Turnitin's own materials on discussing AI writing with students describe the report as something an instructor chooses to share rather than something a student can request on demand [4]. That is the structural gap most students run into: the detector exists, the report exists, but the submission route that produces it belongs to the university, not to the person who wrote the draft [4]. Once a paper is in the institutional system, the report is also attached to that submission record, so experimenting is not free [4].
What students can do is preview the same report before the real submission, using an independent checking service that returns the AI detection report and the similarity report as downloadable PDFs. Turnitin0 runs a non-repository check: the file is analysed without being added to Turnitin's student paper database, English documents between 300 and 30,000 words are accepted, and most orders finish in under 15 minutes. Because the output mirrors what appears in an institutional LMS — including the *% low-confidence display — you see the flagged sentences, not just a number, and you can decide whether a paragraph needs rewriting in your own voice before it ever reaches your department [4].
The value of that preview is diagnostic rather than cosmetic. Turnitin's own guidance to educators is that students benefit from seeing how their writing is read, because the conversation about process is more useful than the score alone [4]. A pre-submission check gives you that same view early enough to act on it — rewriting a flagged passage, adding your own analysis and citations, or keeping notes on how a section was drafted — instead of discovering the highlights after the deadline has passed [4].
If the preview shows highlighted sentences you did not expect, the fastest path forward is to see the full report yourself rather than guess at what a marker will read — turnitin0 returns the actual Turnitin AI and similarity PDFs for your draft, so you can work from the same highlights your instructor would.
※ Turnitin0.com - Actual Turnitin AI Report Cover, Score, Flag And Similarity Summary
FAQ
Is Turnitin's chat detector a separate tool from the similarity checker?
No. AI writing detection runs inside the same Turnitin submission and feedback workflow as the similarity report, so one document produces both an AI percentage with highlights and a similarity summary [1].
Why does my AI report show *% instead of a number?
Turnitin displays *% when the detected AI share falls below its confidence threshold of about 20%, which marks a low-confidence signal rather than a confirmed zero [3].
Does editing a chat-generated paragraph remove the highlight?
Sometimes. The detector reads the text, so rewriting in your own phrasing and adding original analysis can reduce flagged share, but results vary and near the threshold the same passage may score differently on resubmission [2].
Can I see my own AI writing report after I submit?
That depends on your institution. Turnitin's student guidance treats the report as something an instructor chooses to share, so visibility is configured at the institutional level rather than by the student [4].
Will a pre-submission check add my paper to the student database?
Turnitin0 checks without adding the file to Turnitin's student paper database and does not share reports with third-party databases, so a preview does not affect your later official submission.