WGU and Your Grammarly AI Score: Which Number Does an Evaluator See?

If you have a Grammarly AI percentage on one screen and a percentage on your WGU submission page on the other, neither one is a verdict and the two are not measuring the same thing. The WGU number is a similarity score, which WGU defines as "the percentage of matching content in your submission". The Grammarly number is Grammarly's own estimate of how much of your text resembles AI writing. WGU does run AI detection as well, and it has said in writing that students never see that one.

That matters because the Grammarly number is not an input to anything at WGU that I could find in writing. Grammarly itself says its score should not be read as a prediction of what someone else's detector will return, and no WGU page I could read describes a percentage as the trigger for anything.

That is not the same as "nothing is being checked". Something is, and WGU has described it. The rest of this page is that description, plus what the other numbers on your screen are, all of it from pages you can open yourself.

WGU's submission guide is specific about how many results you get: "After uploading your work, refresh the page to see three (3) scan results that may be available to students for review before submission."

The first three are the three columns in the screenshots on WGU's own help page: SIMILARITY, PROFESSIONAL COMMUNICATION, CONTENT. None of them is headed AI.

Which Grammarly surface you are standing on matters too. The free grammarly.com AI Detector page reports "—% of this text appears to be AI-generated" with two sub-labels under it, "Resembles AI text" and "No AI text patterns found". The version documented in Grammarly's support guide is a paid feature that appears inside Google Docs, Word and Grammarly's own editor, and Grammarly for Education plans are on its list of eligible plans. Whether your school has switched that on is a separate question I cannot answer from outside.

Nor does the rest of WGU's submission guide mention AI. I searched that page for "AI" as a whole word: 15,023 characters of article body, zero hits, and zero for "artificial intelligence" too. Counting "the" as a whole word, case-sensitively, the same text returns 117, so the counter was alive. The step-by-step page on submitting a performance assessment does not contain the letters AI as a word anywhere.

Different vendors is the small reason. Different units is the real one.

A similarity percentage rises when your words match text that already exists somewhere. Quote the rubric, restate the task prompt, cite the same nursing standard every other student in the course cites, and it climbs. An AI percentage rises when a model thinks your sentences look machine-written. Neither one moves the other, and there is no conversion between them.

Grammarly says the cross-tool half of this outright, in its AI Detector user guide:

The AI detection feature uses a proprietary in-house model, so the scores may differ from those of other solutions like Turnitin, GPTZero, Copyleaks, and others. You should not view any scores from Grammarly's AI detection as a clear indication that your professor will see the same score when they run it through their preferred AI detector. However, our model is trained similarly to some of the other most accurate and used detectors, so we believe the scores should be directionally aligned.

The last sentence is theirs as well, and it points the other way from the two before it. I have left it in rather than stopping at the convenient clause. Grammarly's position, read whole, is that the scores will not match and that it expects them to lean the same direction.

The same page says the percentage "should not be used as an objective source of truth", and that the feature "is not 100% accurate and should not be used as a definitive assessment of whether AI-generated text is present". Grammarly's public AI Detector page says the tool "Achieves 99% detection accuracy and ranks #1 on RAID's independent benchmark". Both pages are Grammarly's, both were live, and I opened them one after the other in the same session. I am not going to adjudicate that. I am pointing at which sentence sits on which page, so you can weigh whatever number you are holding against both of them.

One line in that guide is worth more than any of the percentages:

traditional, non-generative corrections in the form of Grammarly red and blue underlines should typically not impact the percentage score. However, using more sophisticated Grammarly writing agents in docs, including the Grammarly agent, or using the Grammarly generative AI assistant in Google Docs and Word to meaningfully rewrite sentences and full paragraphs will raise the percentage score.

If your number jumped after an editing pass, that sentence tells you where to look: at whether you accepted underline fixes or agent rewrites. They are not the same input as far as Grammarly's own model is concerned.

This part is on wgu.edu rather than behind the student portal, which is probably why it so rarely comes up. It is a signed blog post, not a policy: "Guiding Responsible Student Use of AI", by Debbie Fowler, Senior Vice President of Academic Delivery, published June 24, 2025.

WGU's similarity checker includes an AI detection tool that operates in the background, collecting data on the presence of AI-generated text in student submissions. This data is accessible only to the academic integrity team for analysis. Although the tool boasts a 0.2% false positive rate, AI detection relies on linguistic patterns—such as sentence lengths, hyphens, average number of syllables, and parts of speech—to identify potential AI-generated content.

The next paragraph gives the decision rule, and it is not a number. It also brings up Grammarly without being asked:

Most detection tools only indicate that content is "likely AI-generated," as they cannot definitively distinguish between AI-generated writing and writing that reflects limited skills. Consequently, human analysis remains essential. Additionally, tools like Grammarly may be flagged due to their influence on sentence structure. Given that even one false positive is unacceptable, WGU relies on overt evidence of generative AI use to identify integrity violations.

Two sentences there, back to back, are worth more than any number on your screen. WGU says tools like Grammarly "may be flagged" for their influence on sentence structure, and then says it relies on "overt evidence of generative AI use" to find violations. So WGU itself says a school-approved grammar tool can show up in a scan, and a scan showing something is not what WGU says it acts on. That is roughly why a student can carry a large Grammarly percentage around for a whole term and never hear from anybody.

That leaves the question of what "overt evidence" means in practice, and the post answers it. Here is what a human evaluator reads for, verbatim, and it is the only published description I have found of what an AI decision at WGU is actually made on:

• Repetitive language
• Grammatical errors
• Vague content
• Lack of personal perspective
• Factual errors
• Fake, irretrievable sources

One of those six is doing most of the work right now. The post calls fabricated citations "a recent, prevalent issue", and it is the only item on the list with a published consequence attached to it:

When unverifiable sources are identified, the evaluation team locks the submission and returns it without evaluation. The student must then meet with their instructor to revise the work, and the instructor must validate all sources before the evaluation is unlocked.

That is the first rung, and the post describes the rest of the ladder in the next paragraph: "egregious integrity violations warrant academic warnings that require revisions", and "Multiple integrity violations result in coaching conversations with the academic coaching center, instruction faculty, and student conduct office." Which is worth knowing if what you are really asking is how worried to be. Watch which word each rung hangs on, though. The academic warning hangs on "egregious", not on repetition, and it is only the conduct office that WGU attaches to "multiple" violations. So a single locked submission is not automatically a conduct matter, and a bad enough first one can still draw a warning.

Two caveats, in the open rather than at the bottom. The post is dated June 2025, while the AI policy took effect on November 1, 2025 and logs revisions up to May 18, 2026. And "boasts a 0.2% false positive rate" is WGU's sentence, WGU's verb, about a tool WGU does not name; the post gives no source for the figure. I have not verified it and neither can you.

The symbol people read as "N/A", a circle with a slash through it, has a documented meaning, and WGU spells it out in a submission FAQ:

I see a message that says, "A scan is not required for your attachment." Is something wrong? No. Some tasks do not require a Similarity or Professional Communication scan. If you see this message, it confirms that a scan is not required; you can safely submit your task. It is not an error, nor does it mean your submission failed. For more help, review Understanding Similarity Reports.

So it marks a scan you are allowed to skip. It is not a result being withheld from you, and it is not an AI verdict of zero either.

Count the icons carefully, though, because the page disagrees with itself. The prose says "The Professional Communication scan will deliver one of four result indicators" and then names four: green checkmark, yellow triangle, the "no" symbol, and the document with an X. The symbol key further down the same page, captioned "Example Professional Communication scan indicator symbol key" and numbered Figure 8, has seven rows. It sits just after the four are listed out, so it is easy to scroll past. The three the prose leaves out are a green thumbs-up ("Your attachment seems ready for evaluation"), a spinner ("Your attachment is being scanned. Please wait 3-5 minutes"), and a red exclamation mark ("There was an error scanning your attachment"). Read only the text and you can easily conclude your icon does not exist.

Here is the one thing I could not check, and it is the part that would help you most. Four WGU articles go further than any of the above, and all four redirect to the student sign-on page: "Understanding Similarity Reports", the "Grammarly for Education FAQ", the "Student Guide to Using AI at Western Governors University", and one that public pages link to as "AI and Academic Integrity at WGU" and whose own address calls it the AI Text Generation in WGU Assessments FAQ. I tried each three times. You can open all four and I cannot, so that is the job I am handing back: read them from inside the portal, and if the icon on your task is still one you cannot place, ask your instructor which of the three scans (similarity, professional communication or content) it belongs to before you rewrite a single paragraph over it.

Both claims get repeated constantly and neither one carries a source, so I went looking for one.

Six WGU pages that bear on this are readable without logging in: the AI policy, the performance assessment submission guide, Upholding Academic Integrity at WGU, the Academic Integrity policy, the Student Code of Conduct, and Assessment Policies. Article body text only, 101,794 characters combined. Across all six:

Same counting rule as earlier, "the" as a whole word and case-sensitive, run across all six bodies: 804. A nonsense control string returns 0. So the zeros above are real zeros and not a broken search.

I could not find any WGU document stating that AI detection has been switched off, and none stating that staff have been told to stop checking. What I did find is the opposite, from WGU itself: a tool that "operates in the background".

That cuts both ways and I would rather say so than watch it get quoted in one direction. Those same six pages contain no sentence confirming detection is switched on either, no published threshold, and no AI percentage. The word "percentage" turns up exactly once in all 101,794 characters, and it is the sentence quoted at the top of this page, the one defining the similarity score. A background tool existing does not mean a number decides anything; a policy page with no threshold on it does not mean nothing is running.

The words "faculty" and "staff" each appear once in the entire AI policy, in the same sentence, and this is it:

Faculty and academic staff may report suspected misuse of AI tools to the academic integrity or student conduct office.

That is permission to report. It says nothing about detectors in either direction. And the count above is a count of what is public, which is not the same as a count of everything WGU has written; the login-walled pages listed in the previous section are outside it.

One piece of this does match something published, and it is probably where the whole thing started. You seeing nothing is exactly what WGU described. The data is "accessible only to the academic integrity team".

Ordered by what WGU has actually put in writing, not by what feels urgent.

1. Open every source and confirm it exists. Of the six patterns WGU says its evaluators read for, fabricated citations is the one it calls "a recent, prevalent issue" and the only one with a published consequence attached: the evaluation team locks the submission and returns it without evaluation, and your instructor has to validate all sources before it reopens. Fifteen minutes clicking through your own reference list is the highest-value thing you can do tonight.
2. Read the similarity report, not the similarity total. Rubric language, the task prompt, and standard definitions you quoted will all match something. WGU's stated requirement is that "your work needs sufficient original content" to be evaluated for competency, which is a question about your paper rather than about a number (how a similarity report differs from an AI report).
3. Check what you accepted in Grammarly, not what it scored. Its own guide treats underline corrections and agent rewrites as different inputs.
4. Check whether your use needs a citation. The policy splits it cleanly. Spelling, grammar, punctuation, and "Enhancing clarity, sentence structure, and fluency of language (e.g., Grammarly for Education, Microsoft Editor)" are listed as usable without citation. "Generating or paraphrasing portions of content", summarizing sources, and "Developing arguments, examples, or supporting details" are on the citation-required list. Course and program instructions can be stricter than the handbook, and the policy says so.
5. Keep the drafting record. It costs nothing and it answers a question no percentage can (version history as evidence).
6. If something is wrong, you will get an email, not a number. WGU's stated process is that "Students are not notified unless a concern is identified", and that the contact comes from an Academic Coaching Center Coach with directions and resources (what to do if you wrote it yourself).

Does WGU use Turnitin? Not in anything I could read. The public pages describe "similarity-checking software" and never name a vendor in the prose, and "Turnitin" does not appear anywhere in the 101,794 characters I searched. Both the submission guide and the academic integrity article carry a "Copyleaks" tag, which is a label WGU attached to the article rather than a sentence WGU wrote. Treat that as a hint, not an answer.

My similarity score is in single digits and Grammarly says over 80%. Which one is wrong? Probably neither. One counts matched text, the other estimates how much of your writing resembles AI output. A paper can honestly be low on one and high on the other, and the gap between them is not evidence about you.

Is Grammarly allowed at WGU? The AI policy names it directly: "Tools that provide basic spelling, grammar, and syntax corrections—such as Grammarly for Education, built-in spell-checkers, and grammar functions in word processing software—are considered acceptable and do not require citation." Generating and paraphrasing sit on the citation-required list instead. Your course instructions can add requirements on top.

Why did my AI score go up when I only fixed the writing? Grammarly's guide separates its red and blue underline corrections, which it says "should typically not impact the percentage score", from its agent rewrites, which it says "will raise the percentage score" because those come from its own LLM.

What does the crossed-out circle or a blank AI check mean? WGU's documented wording is "A scan is not required for your attachment", and its FAQ adds that this "is not an error, nor does it mean your submission failed."

Will the background AI data be used against me? WGU's published position is that the data is "accessible only to the academic integrity team for analysis", that "human analysis remains essential", and that it "relies on overt evidence of generative AI use to identify integrity violations". No WGU page I could read describes a percentage as the trigger for anything.

Neighboring questions this page deliberately leaves alone:

• Does Turnitin detect Grammarly as AI writing? is the reverse case: Grammarly as the thing being scanned, in a course that runs Turnitin. Turnitin has answered it in writing, with a qualifier that most summaries drop.
• Grammarly AI Humanizer review covers Grammarly's separate rewriting tool and what its own page does and does not establish.
• Which writing tools count as "AI assistance" you have to disclose? is the disclosure question across publishers and universities, for when your institution's list runs out before your situation does.
• Why two detectors give the same paragraph different scores goes into the mechanics behind the gap described above, and what AI detectors measure covers what the number is a number of.

Everything above was read logged out on September 15, 2026, in a browser rather than by script, because `cm.wgu.edu` returns 403 to scripted requests and a script would have reported these pages as empty. Institutional pages change. Reopen them before you rely on any of this.

• WGU Student Policy Handbook, Use of Artificial Intelligence (AI) Tools. Effective November 1, 2025; revision log runs to May 18, 2026.
• WGU Knowledge Center, Performance Assessment Submission. Last edited September 8, 2026.
• WGU Knowledge Center, Upholding Academic Integrity at WGU.
• WGU Student Policy Handbook, Academic Integrity, Student Code of Conduct and Assessment Policies. These three plus the two above and the AI policy are the six pages in the 101,794-character count.
• WGU blog, Guiding Responsible Student Use of AI, by Debbie Fowler, SVP Academic Delivery, published June 24, 2025.
• Grammarly Support, AI Detector user guide.
• Grammarly, AI Detector product page.

Behind the WGU student sign-on, and therefore not used here: Understanding Similarity Reports, the Grammarly for Education FAQ, the Student Guide to Using AI at Western Governors University, and the AI Text Generation in WGU Assessments FAQ.