A trusted proofreading tool now uses AI to write for students

If you’re a teacher, you’ve probably heard of Grammarly. Maybe you’ve even recommended that your students use it. After all, the company, which got its start in 2009 as a spelling and grammar checker, boasts more than 40 million users. What you may not know is this: over the past few years, Grammarly has incorporated generative AI in order to write on students’ behalf.

The shift began in earnest in 2019, when Grammarly introduced a tone detector that employed both rules and machine learning to change the emotional tenor of a piece of text. The feature began with email, then expanded to other kinds of writing.

After that, Grammarly continued to tie its future even more closely to AI. In 2020, Grammarly made its first investment in another company. The lucky recipient was Docugami, which uses AI to help clients process documents. Three years later, as ChatGPT took the world by storm, Grammarly leaned into the technology. The company utilized OpenAI’s large language models to launch GrammarlyGo, which allowed users to enter prompts to generate ideas and text.

Grammarly’s homepage now advertises “Responsible AI that ensures your writing and reputation shine,” promising students an “an AI writing partner that helps you find the words you need⁠.” Students can now take advantage of Grammarly’s AI capabilities to rewrite paragraphs or even write essays for them, exactly the same way they might with ChatGPT.

For teachers who still think of Grammarly as a more advanced version of autocorrect, this can be disastrous. While you may be okay with software that teaches students how to spell or how to avoid the passive voice, you want to know that they’re learning to express their own ideas, understanding information well enough to paraphrase it themselves, considering their word choice, and revising their own work,

The good news is that Pangram will pick up on uses of Grammarly that stray too far from simple editing. If your students rely on Grammarly’s generative AI features, we’ll let you know.

Grammarly’s latest technological leaps have made it clear that teachers need a reliable way to determine if students are using it inappropriately. Last February, for instance, The New York Post reported that a Georgia college student was put on academic probation after submitting a paper that triggered an AI detector. The student insisted she had only used the free version of Grammarly to proofread the assignment, and the company provided a statement saying that its basic suggestions “are not powered by generative AI” and that some software could incorrectly flag them.

Other schools are working to avoid such thorny situations altogether. After repeated issues with submissions from students who said they had used Grammarly, Notre Dame updated its policy in August to clarify that professors who prohibit generative AI are also prohibiting editing tools like Grammarly.

Grammarly itself seems to understand the potential for students to misuse its new capabilities. The company has recently rolled out “authorship” features that keep a record of a student’s writing process, distinguishing between typed text, copy-pasted paragraphs, and AI-generated language. Students can then send the report of their writing process to their instructors as proof that their use of Grammarly was within the bounds of academic integrity if their writing triggers AI detection.

Now that you know the difference between the Grammarly of today and the one you knew a decade ago, you should treat it like any other AI tool. Clarify to your students whether and how they are allowed to use it and if you have any questions about how we evaluate text that has been edited with Grammarly, reach out at info@pangram.com.