AI-Assisted vs AI-Generated: How Turnitin Treats Each One

Turnitin's AI detection does not track your writing process. It does not know whether you drafted by hand, dictated, or used a tool at some stage. It evaluates the text you submit.

"When a paper is submitted to Turnitin, sentences from the submission are extracted and segmented into overlapping sections for prediction analysis. Each segment is classified by the AI detection model and given a value between 0 and 1, denoting the probability of the text being likely human or AI-generated."

The model extracts sentences from the finished document, segments them, and scores each segment. It does not have access to version history, editing timestamps, or tool usage logs, which is exactly why that record is worth keeping on your side, as how to prove you didn't use AI sets out. It works with what is on the page at the moment of submission.

This means the distinction between "I wrote this with help" and "this was generated for me" is not something the model can observe directly. It classifies text based on statistical patterns in the final output.

The scope of what gets analyzed is also specific: "This qualifying text includes only prose sentences, meaning that we only analyze blocks of text that are written in standard grammatical sentences and do not include other types of writing such as lists, bullet points (short non-sentence structures), or other non-sentence structures." And "This percentage is not necessarily the percentage of the entire submission."

Many students use Grammarly or similar tools to fix spelling and grammar. The question is whether those corrections cause the text to be flagged as AI-generated. Turnitin's documentation addresses this directly.

"Our detector is not tuned to target Grammarly-generated spelling, grammar, and punctuation modifications to content but rather, other AI content written by LLMs such as GPT-3.5. Based on tests we conducted on human-written documents with no AI-generated content in them, in most cases, changes made by Grammarly (free & premium) and/or other grammar-checking tools were not flagged as AI-written by our detector."

The key phrase is "in most cases." The detector is not tuned to catch spelling and grammar corrections. In most cases, text that a human wrote and then cleaned up with Grammarly will not be flagged. This applies to both the free and premium versions of Grammarly, and to other grammar-checking tools as well. Does Turnitin detect Grammarly as AI writing goes through the vendor wording line by line.

This does not mean grammar-edited text is invisible to the detector. It means the detector is not designed to identify that specific type of modification. The finished prose still gets analyzed for AI-like statistical patterns. But the act of running Grammarly for spelling and grammar does not, in most cases, cause a flag on its own.

The documentation draws a clear line between Grammarly's correction features and its generative features. The exclusion does not extend to content produced by generative AI tools.

"Please note that this excludes content generated by Grammarly's generative AI-powered features, including draft generation, paraphrasing, summarizing, and other features. Content produced using these features will likely be flagged as AI-generated by our detector."

So there are two categories. On one side: spelling, grammar, and punctuation corrections. These are not targeted by the detector, and in most cases are not flagged. On the other side: draft generation, paraphrasing, summarizing, and other generative features. Content produced using these features will likely be flagged as AI-generated.

The distinction matters. If you used Grammarly to fix typos and tighten grammar, the detector is not tuned to catch that. If you used Grammarly to generate a draft, paraphrase a passage, or summarize a source, that output is treated like any other AI-generated content and will likely be flagged.

This line is based on what the features produce, not on the brand. A generative feature from any tool, not just Grammarly, falls on the "will likely be flagged" side. A spelling or grammar correction from any tool falls on the "not tuned to target" side. Which side you land on also decides what you have to say about it, covered in does running your document through a rewriting tool change what you have to declare.

Even when no AI was involved, certain types of human-written prose can trigger false positives. The documentation lists specific patterns to watch for.

"Sometimes false positives (incorrectly flagging human-written text as AI-generated), can include content without a lot of structural variation, text that literally repeats itself, or text that has been paraphrased without developing new ideas."

"If our indicator shows a higher amount of AI writing in such text, we advise you to take that into consideration when looking at the percentage indicated."

So if your writing is structurally flat, repeats the same phrasing, or paraphrases without adding new ideas, the model may flag it even if it is entirely human-written. This is relevant to the AI-assisted distinction because a student who writes their own text but paraphrases heavily could see a flag that has nothing to do with using a tool at all. I paraphrased AI text with another AI and Turnitin still flagged it is the other end of the same effect.

Turnitin's documentation is explicit that the AI detection score should not be treated as definitive proof of misconduct.

"Our AI writing detection model may not always be accurate (it may misidentify human-written, AI-generated, and AI-paraphrased text), so it should not be used as the sole basis for adverse actions against a student."

"It takes further scrutiny and human judgment in conjunction with an organization's application of its specific academic policies to determine whether academic misconduct has occurred."

The score is one piece of information. It requires further scrutiny and human judgment, applied in the context of the institution's specific academic policies. The model can misidentify text in both directions: flagging human text as AI, or missing AI text entirely. Treating the percentage as a sole basis would ignore the documentation's own guidance. If you are the one holding the flagged report, flagged, but you wrote it yourself covers how to put that in front of a marker.

If you are trying to understand how Turnitin treats AI-assisted writing versus AI-generated writing, here is the summary:

• The model sees the finished text. It does not track your writing process. It segments and scores prose sentences from the submitted document.
• Spelling and grammar edits are different. The detector is not tuned to target Grammarly-style spelling, grammar, and punctuation corrections. In most cases, these are not flagged.
• Generative features will likely be flagged. Content from Grammarly's draft generation, paraphrasing, summarizing, and other generative features is treated as AI-generated and will likely be flagged.
• False positives happen. Structurally flat, repetitive, or paraphrased prose is more likely to be flagged, even with no AI involvement.
• The score is not a sole basis. The documentation says it should not be used as the sole basis for adverse actions against a student. Human judgment and institutional policy apply.

Eligible passages can be re-run at no charge.