Of course, no one learns to be a better writer when someone else (or some AI bot) is doing the work for them. The question is whether chatbots can morph into decent writing teachers or coaches that students actually want to consult to improve their writing, and not just use for shortcuts.
Maybe.
Jennifer Meyer, an assistant professor at the University of Vienna in Austria, has been studying how AI bots can be used to improve student writing for several years. In an interview with The Hechinger Report, she explained why she is cautious about the ability of AI to make us better writers and is still testing how to use the new technology effectively.
Meyer says that just because ChatGPT is available 24/7 doesn’t mean students should consult it at the start of the writing process. Instead, Meyer believes that students would generally learn more if they wrote a first draft on their own.
That’s when AI could be most helpful, she thinks. With some prompting, a chatbot could provide immediate writing feedback targeted to each student’s needs. One student might need to practice writing shorter sentences. Another might be struggling with story structure and outlining. AI could theoretically meet an entire classroom’s individual needs faster than a human teacher.
In Meyer’s experiments, she inserted AI only after the first draft was done as part of the revision process. In a study published in 2024, she randomly assigned 200 German high school students to receive AI feedback after writing a draft of an essay in English. Their revised essays were stronger than those of 250 students who were also told to revise, but didn’t get help from AI.
In surveys, those with AI feedback also said they felt more motivated to rewrite than those who didn’t get feedback. That motivation is critical. Often, students aren’t in the mood to rewrite, and without revisions, students can’t become better writers.
Meyer doesn’t consider her experiment proof that AI is a great writing teacher. She didn’t compare it with how student writing improved after human feedback. Her experiment compared only AI feedback with no feedback.
Most importantly, one dose of AI writing feedback wasn’t enough to elevate students’ writing skills. On a second, fresh essay topic, the students who had previously received AI feedback didn’t write any better than the students who hadn’t been helped by AI.
It’s unclear how many rounds of AI feedback it would take to boost a student’s writing skills more permanently, not just help revise the essay at hand.
And Meyer doesn’t know whether a student would want to keep discussing writing with an AI bot over and over again. Maybe students were willing to engage with it in this experiment because it was a novelty, but could soon tire of it. That’s next on Meyer’s research agenda.
A much smaller MIT study published earlier this year echoes Meyer’s theory. “Your Brain on ChatGPT” went viral because it seemed to say that using ChatGPT to help write an essay made students’ brains less engaged. Researchers found that students who wrote an essay without any online tools had stronger brain connectivity and activity than students who used AI or consulted Google to search for source materials. (Using Google while writing wasn’t nearly as bad for the brain as AI.)
Although those results made headlines, there was more to the experiment. The students who initially wrote an essay on their own were later given ChatGPT to help improve their essays. That switch to ChatGPT boosted brain activity, in contrast to what the neuroscientists found during the initial writing process.
These studies add to the evidence that delaying AI a bit, after some initial thinking and drafting, could be a sweet spot in learning. That’s something researchers need to test more.
Still, Meyer remains concerned about giving AI tools to very weak writers and to young children who haven’t developed basic writing skills. “This could be a real problem,” said Meyer. “It could be detrimental to use these tools too early.”
Meyer doesn’t think it’s always a bad idea for students to ask ChatGPT to do the writing for them.
Just as young artists learn to paint by copying masterpieces in museums, students might learn to write better by copying good writing. (The late great New Yorker editor John Bennet taught Jill to write this way. He called it “copy work” and he encouraged his journalism students to do it every week by copying longhand the words of legendary writers, not AI.)
Meyer suggests that students ask ChatGPT to write a sample essay that meets their teacher’s assignment and grading criteria. The next step is key. If students pretend it’s their own piece and submit it, that’s cheating. They’ve also offloaded cognitive work to technology and haven’t learned anything.
But the AI essay can be an effective teaching tool, in theory, if students study the arguments, organizational structure, sentence construction and vocabulary before writing a new draft in their own words. Ideally, the next assignment should be better if students have learned through that analysis and internalized the style and techniques of the model essay, Meyer said.
“My hypothesis would be as long as there’s cognitive effort with it, as long as there’s a lot of time on task and like critical thinking about the output, then it should be fine,” said Meyer.
Everyone likes a compliment. But too much praise can drown learning just as too much water can keep flowers from blooming.
ChatGPT has a tendency to pour the praise on thick and often begins with banal flattery, like “Great job!” even when a student’s writing needs a lot of work. In Meyer’s test of whether AI feedback can improve students’ writing, she intentionally told ChatGPT not to start with praise and instead go straight to constructive criticism.
Her parsimonious approach to praise was inspired by a 2023 writing study about what motivates students to revise. The study found that when teachers started off with general praise, students were left with the false impression that their work was already good enough, so they didn’t put in the extra effort to rewrite.
In Meyer’s experiment, the praise-free feedback was effective in getting students to revise and improve their essays. But she didn’t set up a direct competition between the two approaches — praise-free vs. praise-full — so we don’t know for sure which is more effective when students are interacting with AI.
Being stingy with praise rubs real teachers the wrong way. After Meyer removed praise from the feedback, teachers told her they wanted to restore it. “They wondered about why the feedback was so negative,” Meyer said. “That’s not how they would do it.”
Meyer and other researchers may one day solve the puzzle of how to turn AI chatbots into great writing coaches. But whether students will have the willpower or desire to forgo an instantly written essay is another matter. As long as ChatGPT continues to allow students to take the easy way out, it’s human nature to do so.
This story was produced by The Hechinger Report, a nonprofit, independent news organization focused on inequality and innovation in education, and reviewed and distributed by Stacker.