Aboli Karnik You are correct - the educator is the best judge. Your knowledge of the student's voice, writing habits and context of the assignment is the most reliable form of assessment.

AI tools primarily look for patterns, sentence structure and vocabulary choices common to large language models (LLMs), and a student's original work can sometimes mimic the output of an LLM. Additionally, built-in AI-powered features within Grammarly, MS Word, Google Docs, etc. that "refine" or "improve" text are also widely used by students during the editing process. These tools can introduce AI-generated text in ways that may not be obvious to users. Considering these points, detectors should be used to provide instructors with data signals, but they should never be used over professional judgment. 

Aboli Karnik I want to add my thoughts to Gailene's response. As a former educator myself, one of the things we're proud of at Turnitin is that we have a whole swath of former educators built in across many different functions at the company, including on our AI team. I point that out because one of our grounding philosophies is to respect educators for their professional judgment and discretion. It is never our intention to supplant the educator; it is only ever our intent to supplement with the data points and insights our tools can provide. 

Another area of your original question worth considering is HOW educators and students can demonstrate that the work is the student's original work. As a former English teacher, I believe that some of the core tenets of writing pedagogy remain true, and one of those is around the writing process. Using a process approach to writing assignments helps to demonstrate how the work takes shape over time, where educators have some visibility into the choices the student writer has made along the way, including whether they did/did not use generative AI and how they used it if they did. My team of pedagogical experts has been saying this since even before Turnitin's AI detector launched, and we're still saying it. Putting that kind of visibility together with a detector is a way of bringing together multiple data points.

With the release of Turnitin Clarity in 2025, there's now even more visibility possible as Turnitin Clarity can make the writing process even more transparent. We often talk about how Turnitin Clarity helps educators, but in this case, it helps students as well. Students told us that they feared not being able to prove or document their integrity, and tools like Turnitin Clarity give them an actual record of how their writing took shape. We like to say that its use can help to rebuild any erosion of trust between educator and student. 

Mary Beth Kwase Hi Mary Beth! Thank you for your detailed feedback! There's a lot to unpack, but we're looking into each of your points. It may take us a few days to get back to you, with the U.S. holidays, but I wanted you to know we're looking!

Hi Mary Beth Kwase. 

We really appreciate your early adoption of Clarity, and look forward to working with you and other educators as we build out new functionality that supports a more transparent and formative learning process.

I believe your first set of feedback is related to the AI Writing report, but if that’s not the view you are referring to, please let me know. In this case, our team is actively investigating how we can distinguish the AI contributed from the Clarity Assistant from other AI content. We are looking to solve for this in 2026.

It’s also important to note that many common tools like grammar and spelling - even auto-correction - are powered more and more by AI now, so using those tools can also contribute to the detection score. That said, I’d be remiss if I didn’t emphasize that our detection tools do not make any determination of misconduct. We want you to provide you with data that contributes to how you make your decisions about academic integrity. Detection is just one data point in your process. 

I’d like to dive into the second part a little more. Are you looking for more grading tools to use as part of the writing process (pre-submission), or as part of the Clarity Writing Report? Apologies if you already have this information, but we do have advanced grading tools as part of the final paper Similarity Report view (more details here).

Aside from being able to highlight and add comments, what other tools are you using via your Canvas integration that you would like to see when grading the students’ Clarity Writing Report? I’m also curious to understand how we might improve the grading and feedback discovery and flow between the various reports.

Thank you again for your patience with our response, and for sharing your experience with Clarity. I know this is a busy time of year, so I’ll keep an eye on this thread for any new responses.

Best -

Gailene

Mary Beth Kwase I appreciate your persistence! It's really helpful to understand how you would like to use these tools, and how you are finding your way. I would love to connect you with one of our Product team members so you can walk them through your experience.

If I understand what you are saying about the AI Assistant, it's giving you way more content than you would have expected. I suspect that different levels of complexity with different prompts will have some variability, but our goal has never been to write the work for the student. 

As you know, we just launched Clarity this year, and the team is actively working to address enhancements and updates as we hear back from our early adopters, like you. We are in the early stages of defining AI customization features that will allow instructors to configure the type of assistance they want to enable (or restrict) within the AI Chat for students. I've passed your feedback along to the team so they can take this into consideration. There's definitely room to grow and improve our user experience, so feedback with context like yours really helps - thank you!

Michael Augello 

Hi Michael — really appreciate you joining our chat and bringing such a meaty question. I want to say up front: you’re asking the exact questions I hear from so many educators right now. Before I stepped into my role here as Sr Director of Customer Engagement at Turnitin, I taught for years too, and the feeling you described — trusting your professional judgment but also second-guessing more than you ever expected in year 12 — is so real. You’re not alone!

Let's unpack a couple things from your post:

First: the uptick in AI scores.
You’re not imagining it. We updated our AI detector recently to better identify content from the newest large language models (think GPT-5, Gemini 2.5 Flash, etc.) and AI bypassers. Anytime generative AI advances, we have to move just as quickly to keep detection accurate. So the higher scores you’re seeing lately? That tracks with what we’d expect after an update.

We test extensively before we release anything — more than 700,000 academic papers written before ChatGPT even existed, plus another nearly one million human-written documents based on customer feedback. That’s how we maintain a false positive rate below 1%. In our latest testing, even among documents showing more than 20% AFTER the update, that false positive rate stayed under 1%.

All that said, the AI score is still just one data point. It should never be the sole basis for an academic integrity decision, and you’re already doing exactly what we encourage educators to do: talking with students, reviewing drafts, and leaning on your knowledge of your students and their writing.

On Grammarly, Gemini, and other tools:
You hit the nail on the head. Students often don’t realize that when they accept extensive revision suggestions — not just punctuation fixes, but full-sentence replacements or paraphrasing — those changes DO introduce AI-generated content. And that can absolutely raise the AI score. It doesn’t automatically mean its misuse; it just means their writing now contains patterns common in AI-generated text, which you, as the educator can now identify and decide how to approach instructionally.

That’s can be helpful framing for students:
“Your score may be higher not because you did something wrong intentionally, but because the tool changed your voice more than you realized. Now, what do you want to do about it?”

Your initial post is so packed with ideas that I want to bring in another voice from our team. I’m tagging in another former educator at Turnitin, who helps create many of our pedagogical resources, to address false positives. Karen - Turnitin Team Your turn!