Does Grammarly Count as AI? What Turnitin Sees and Schools Allow

Grammarly's spelling, grammar and punctuation corrections are not generative AI, and Turnitin says in its own FAQ that its detector isn't tuned to flag them. Grammarly's Rewrite, Paraphrase, "Improve it" and "Write with generative AI" features are generative AI: Turnitin says their output "will likely be flagged," and most university policies treat them the same as ChatGPT. The grey zone is the middle: accepting a lot of Premium clarity and tone rewrites keeps your ideas but flattens your phrasing, which is exactly the pattern detectors score as AI-like.

Which Grammarly features count as AI

"Grammarly" is now a bundle of very different tools under one name, and many policies don't distinguish them.

The rule of thumb: if Grammarly changed a word you already wrote, you used a checker. If Grammarly wrote a sentence you didn't, you used generative AI.

Does Grammarly count as AI in Turnitin?

Under "If students use Grammarly for grammar checks, does Turnitin detect it and flag it as AI?" Turnitin's AI writing detection FAQ says no: the detector is "not tuned to target Grammarly-generated spelling, grammar, and punctuation modifications," and in tests on human-written documents, changes made by Grammarly free and premium "were not flagged as AI-written" in most cases. The same answer carves out Grammarly's generative features, "including draft generation, paraphrasing, summarizing," whose output "will likely be flagged."

Three other details from that FAQ matter for Grammarly users:

• Turnitin keeps its false positive rate under 1% only for documents it scores above 20% AI. Scores from 1% to 19% are shown as an asterisk with no highlights because false positives cluster there. A lightly Grammarly-edited paper that picks up a few AI-like sentences lands in that asterisk zone, not at 60%.
• Turnitin lists the kinds of human text that produce false positives: "content without a lot of structural variation, text that literally repeats itself, or text that has been paraphrased without developing new ideas." That last phrase describes what heavy rewrite acceptance does to an essay.
• The detector also looks for AI text that was run through a paraphraser or "bypasser," and Grammarly's Paraphraser is a paraphraser. For how that detection behaves, see can Turnitin detect QuillBot paraphrasing.

Why heavy editing moves the score when light editing doesn't

Turnitin's FAQ explains its model in one sentence: language models pick "the next word in a sequence of words in a consistent and highly probable fashion," while human writing is "inconsistent and idiosyncratic." Fixing "recieve" doesn't change how probable your word choices are. Accepting thirty clarity rewrites does: each one swaps your slightly odd, specific sentence for the most expected version of it, and a whole paper of those reads like the average of everyone's writing, which is what the detector was trained to call AI.

Two tests show the split. Originality.ai ran ten human-written IELTS essays through Grammarly twice: accepting only grammar suggestions left almost all of them reading as human, while accepting every rephrase and rewrite suggestion "had a significant impact" on how many were flagged. A University of Lethbridge team published a small peer-reviewed study in 2025 after Turnitin gave a human-written abstract the group had edited with Grammarly an AI score of 16%; in their controlled runs, Grammarly Premium versions of three pre-AI texts mostly scored 0% on Turnitin, with one at 11%. All of those numbers would show as an asterisk under Turnitin's current display. Light Grammarly use barely registers; the problems start when you let it rewrite.

Some institutions take a harder line than Turnitin. Notre Dame's honor code office says "even a small number of suggestions from an AI editing tool" can get a paper flagged, and calls that correct rather than a false positive because the suggestion itself was AI-generated. Turnitin's tests say otherwise for basic corrections, but if you're at Notre Dame, the honor code governs you, not Turnitin's FAQ.

Is using Grammarly cheating in college?

At most universities the instructor sets the rule per course, sometimes per assignment. What is consistent is the distinction schools draw. Four verified examples:

• George Mason University: spell-check and grammar-check features "are not prohibited," but "using the generative AI portions of Grammarly or other software to do the academic work may still be considered a violation." Source
• University at Buffalo (Office of Academic Integrity): "it is often acceptable to use Grammarly (or similar tools) to correct what you have written, it is usually NOT acceptable to have these tools re-write/re-generate text." Source
• University of Notre Dame (Fall 2024 policy update): AI-powered editing tools like Grammarly and WordTune produce AI-generated suggestions, so "if your instructor prohibits the use of gen AI on an assignment or in a class, this prohibition includes the use of editing tools, unless explicitly stated otherwise." Source
• University of Birmingham (Academic Skills Centre): allowed are spelling, punctuation and basic grammar fixes and shortening sentences without changing meaning; not allowed are rewriting for tone or academic style, generating new content, or refining an argument, which the Code of Practice says "may be deemed to be plagiarism." Source

None of these say "Grammarly is fine" or "Grammarly is banned." The line is the same everywhere: correcting what you wrote versus having the tool write.

Is Grammarly Premium cheating?

The subscription isn't the issue. Per Grammarly's plans page, Pro adds "rewrite full sentences with a click," tone adjustment, the plagiarism and AI detectors, and 2,000 generative prompts. Accept only the red underlines and you've done nothing a free user couldn't do. Accept the sentence rewrites and tone changes in a course that bans generative AI and you've used generative AI. The risk is in the suggestions, not the price.

Is using Grammarly Rewrite or Paraphrase cheating?

In a no-AI course, yes, and expect both the policy and the detector to treat it that way. In a course that allows AI with disclosure, it's allowed if you disclose it, in the same words you'd use for a ChatGPT rewrite.

What happened to Marley Stevens at UNG

Stevens, then a junior at the University of North Georgia, submitted a criminal justice paper on October 3, 2023 and was told on October 26 that Turnitin had flagged it as AI-written; she received a zero. The syllabus said not to use AI; she said she used the Grammarly browser extension for spelling and punctuation and hadn't thought of it as AI. The zero cost her HOPE Scholarship eligibility, and after a February 16, 2024 hearing she was placed on academic probation for a year and required to complete a $105 integrity seminar. She was told the outcome couldn't be appealed because it didn't involve suspension or expulsion. EdSurge and Her Campus have the fullest accounts.

What followed: her TikTok passed 5 million views, Grammarly contributed $4,000 to her GoFundMe (the value of the lost scholarship) and later hired her for sponsored videos, and UNG briefly pulled Grammarly from its recommended-resources page. Turnitin's chief product officer told EdSurge that traditional grammar-checking tools don't set off its alarms. Stevens says her case helped push Grammarly to build Authorship; she graduated in December 2025 with the violation still on record. The episode picked up the nickname "Grammarlygate" from a widely shared essay arguing that the real failure was administrators trusting a score over judgment.

The lessons: the syllabus said no AI and the student didn't know her tool counted; there was no process evidence to put beside the detector score; and a single number decided the outcome. If you're ever in that position, start with what to do when your essay is flagged as AI but you wrote it.

How to use Grammarly safely in a graded course

1. Read the AI section of every syllabus and save a screenshot. It's the first thing an integrity office will ask about.
2. If it doesn't mention editing tools, ask before the first assignment, in writing. Keep the reply.
3. Accept corrections, not rewrites. Red underlines are fine. When a suggestion replaces a whole sentence, retype the fix in your own words if you agree with it, or dismiss it.
4. Don't touch Rewrite, Improve it, Paraphraser, tone adjustment, or AI Chat on graded work unless generative AI is explicitly permitted.
5. Turn on Grammarly Authorship before you type the first word. It's off by default and can't reconstruct what happened before you enabled it.
6. Write in a document with version history and avoid pasting in large blocks.
7. Disclose in a line at the end if any AI-assisted feature was used and allowed.
8. Run the final draft through an AI checker and read what it highlights. If heavy editing pushed a section toward AI-like scores, restore your own phrasing there. Don't run it through another paraphraser; that is the "paraphrased without developing new ideas" pattern Turnitin names.

How to ask your instructor

Subject: Quick question about Grammarly and the AI policy

Hi Professor Lee, your syllabus says generative AI isn't allowed on the essays. I use the Grammarly extension for spelling and grammar corrections only, with its rewrite and generative features off. Is that acceptable under the policy, and would you like me to note it on submissions or attach a Grammarly Authorship report? Thanks, Dana

Most instructors will say yes to corrections; the ones who won't will tell you now rather than at a hearing.

How to disclose

If disclosure is required, one line under your references is enough:

I used Grammarly for spelling and grammar corrections. I did not use its rewrite, paraphrase, or text-generation features. Authorship report: [link].

If you used an allowed generative feature, name it and where: "Grammarly's tone suggestions were applied to the introduction; the content and argument are my own."

Grammarly Authorship: what it records and whether it counts as evidence

Authorship is the feature Grammarly built after the false-flag stories. Per Grammarly's support documentation, it labels text as you write it: typed, pasted from a website, pasted from an unknown source, generated with an AI tool such as ChatGPT, or edited with Grammarly's own suggestions. The report has analytics (time on document, sessions, percentage typed vs sourced), a color-coded copy of the text, and a replay of the session showing text written and deleted, sites copied from, and prompts sent to AI tools. It's free, works in Google Docs and Word, and nothing is sent to anyone unless you share a link.

Since September 2025 instructors on Canvas can tick a "Require Authorship Report" box on an assignment, and the report attaches to the submission automatically when the student has it on, per Instructure's announcement. Two limits: it's off by default, so a report that starts halfway through a draft proves less; and no integrity office is obliged to treat it as conclusive (Grammarly itself tells faculty not to use it as a single data point). Treat it as the strongest piece of a package that also includes version history and notes; see how to prove you didn't use AI for the rest.

What's an acceptable score on Grammarly's own AI detector?

Grammarly doesn't say. Its AI Detector user guide calls the percentage "an averaged estimate of the amount of AI-generated text," says the tool "is not 100% accurate and should not be used as a definitive assessment," and adds that Grammarly's own rewriting agents will trigger it, since those rewrites come from its model.

Read it the way you'd read any detector: a high number on a section you wrote means the section reads as uniform and predictable, not that you cheated. The fix is to put your specific phrasing back. For how much weight to give any percentage, see how accurate Turnitin's AI detection is.

A before-and-after

A paragraph after every clarity and tone suggestion was accepted:

Social media has a significant impact on adolescent mental health. Numerous studies have demonstrated that excessive use is associated with increased anxiety and depression. It is essential for parents and educators to implement effective strategies to mitigate these effects.

Same shape every sentence, generic verbs, no trace of the writer. The earlier version, with only the spelling fixed:

My cousin deleted Instagram for a month last year and slept better within a week. That's one kid, but the bigger surveys point the same way: the more hours teens spend scrolling, the more anxiety they report. Parents can't ban phones, so the question is what actually helps.

Uneven sentences, a concrete detail, an opinion. Paste both into Rewritica's free AI checker and compare; the second usually scores far lower, and it wasn't written with any tool at all.

Frequently asked questions

Does Grammarly count as AI in Turnitin?

Not the basic part. Turnitin's own FAQ says its detector isn't tuned to flag Grammarly's spelling, grammar and punctuation changes, and that in its tests those edits were mostly not flagged. It also says text produced by Grammarly's generative features (draft generation, paraphrasing, summarizing) will likely be flagged.

Is using Grammarly cheating in college?

Using it to correct your own sentences is allowed at most schools and in most courses. Using Rewrite, Paraphrase, Improve it, or Write with generative AI is generative AI use, and it's a violation wherever generative AI is banned. When the syllabus doesn't say, ask, because the instructor decides.

Is Grammarly Premium cheating?

The plan isn't what matters; the buttons you press are. Pro adds full-sentence rewrites, tone adjustment and 2,000 generative prompts. Accepting grammar fixes on Pro is the same as on Free. Accepting rewrites or generated text is generative AI use under most policies.

Does Grammarly's paraphrase or rewrite count as AI?

Yes. Grammarly describes its Paraphraser as AI-powered, Turnitin says paraphrasing output will likely be flagged, and Notre Dame's honor code treats editing-tool suggestions as AI-generated text. Treat these features exactly like a ChatGPT rewrite: allowed only if generative AI is allowed, and disclosed.

What is an acceptable AI score on Grammarly's own detector?

Grammarly doesn't publish one. Its guide calls the percentage an averaged estimate that shouldn't be a definitive assessment, and notes that its own rewriting agents will trigger it. Use the score to find stretches that read as machine-made, then restore your own phrasing; don't chase 0%.

Can Grammarly Authorship prove I wrote my essay?

It can show your writing process: typed, pasted, AI-generated and Grammarly-edited text, plus a replay of the session. It's free and works in Google Docs, Word and Canvas, but only if you turn it on before you start. It's strong supporting evidence, not a guarantee any instructor must accept.

Bottom line

Grammarly's red underlines are a spellchecker and nobody serious, including Turnitin, treats them as AI writing. Its rewrite, paraphrase and drafting features are generative AI, governed by whatever your course says about ChatGPT. The trouble is in between, when a student accepts dozens of Premium rewrites in a no-AI course and the paper comes out sounding like everyone's paper. Before the next assignment: read the AI clause, ask about editing tools if it's silent, switch Authorship on, accept corrections only, and disclose anything beyond that. If a checker still flags a section you wrote, rewrite it in your own voice rather than feeding it to another tool.