AI Is Supercharging Your Top Students, And Probably Harming the Rest
A small fraction of high-achievers may well use AI to accelerate their academic growth. What will become of everyone else?
Your content has been saved!
Go to My Saved Content.Chatbots like Claude, AI-powered writing apps like Grammarly, and personal assistants like Meta’s Muse are already a steady part of the teenage diet. According to multiple reports, a majority of teens and tweens regularly use AI—asking it for advice, posing intimate questions, or even befriending the software. No topic appears to be too personal.
When schoolwork is on the table, nearly 90 percent of students turn to AI for help, reports Common Sense Media. The majority say they could easily go without services like Google’s Gemini for an entire month, but 16 percent admit that the technology has gained the upper hand: they struggle to “start or complete their homework” when they can’t use AI.
That’s likely because learning is strenuous and effortful, and the human brain is “designed to be efficient,” writes instructional coach Tyler Rablin. To conserve energy, it’s “constantly looking for the most energy-efficient route to a goal.” When presented with a tool that can calculate the volume and surface area of a cylinder or generate a superficially impeccable essay in seconds, the path of least resistance is virtually impossible to ignore.
“Sure,” cognitive scientist Daniel Willingham concedes on his Substack, some students could learn from AI by “asking it to critique their work” or “to generate counterarguments.” And some do use the tech to probe their understanding or surface gaps in their knowledge.
Take 15-year-old Arjun Sharda, who uses ChatGPT “to make myself practice sample test questions that I will answer with my own work,” he tells NPR. Or consider high school freshman Mason Durgan, who prefers an app called Quiz AI: “It gives you step-by-step tutorials instead of just giving you the answer. I try to limit getting the straight answer as much as I can.” Superintendent of Spokane Public Schools Adam Swinyard tells The Spokesman Review that his daughter turned her biology notes into an AI-generated podcast to listen to while going for a run.
But that isn’t how most students are using the technology, and therein lies one paradox of generative AI: “A small sliver” of students may use it to amplify their learning, Rebecca Winthrop, director of the Center for Universal Education at Brookings, told me in a recent interview. But they are the exception. For the “vast majority,” she said, AI acts as a demotivating force, spurring kids to ask why the work they are assigned is worth their time and effort, especially if they think an LLM can do that work better and faster.
For Winthrop, the conversation around what role AI should or should not play in K-12 education is a question of equity. “I have been thinking more and more about what I see on the horizon,” she warned in a recent issue of her newsletter—“an increasing divergence between students who are highly motivated and engaged and those who are not… What feels different now is that AI may amplify these differences.”
This great divide will not be defined by who uses AI and who doesn’t, clarifies journalist Jenny Anderson on her Substack. The chasm will divide those who “care about getting smarter, digging in, pushing through hard things, grappling with uncertainty” from those who are content to “coast along, resist hard work, or blindly jump through hoops put up for them.”
A TAXONOMY OF DISENGAGEMENT
In 2025, Winthrop and Anderson published their book The Disengaged Teen, a sobering analysis of why students detach from school and the role families can play in reigniting a drive to learn.
Echoing what classroom teachers have long reported, this was not a garden-variety teenage apathy that families could afford to ignore. Relying on over 100 interviews with teenagers and a survey of over 65,000 third through 12th graders conducted by Winthrop and Transcend Education, the two writers proposed a novel taxonomy of academic engagement, describing four distinct modes that students move between throughout the school day, depending on a number of variables.
Roughly half of middle and high schoolers are coasting along in Passenger mode, Winthrop and Anderson say. Those students are showing up for school every day, but drifting along “doing the bare minimum,” spectators in the academic arena who “sometimes bring home high grades, but never fully engage in their work.” Another smaller but sizable fraction, approximately 15 percent, are in Resister mode, overtly refusing to do the work, disrupting class for other students, or often just not showing up at all.
Though only about 4 percent of students across grades 6-12 reported experiences in school that are likely to support Explorer Mode—a state where they feel “deeply involved and engaged in their learning, finding meaning amid the hard work”—a full 30 percent of students in Achiever mode are also “expending tremendous energy doing well in school, getting top marks on exams and studying for hours on end.”
While this group of deeply engaged high-achievers can use tools like ChatGPT and Gemini to pressure-test their understanding of plate tectonics or find flaws in their writing, the same tools feel problematic in the hands of Resisters and Passengers, who together make up 80 percent of middle and high school students. Tweens and teens who are ambivalent about an academic topic—or completely disinterested—now have unfettered access to software that will complete the task for them: A worksheet’s worth of math problems solved instantly, or a pristine paragraph of text written in the target language for your French or German class. Teachers do not have the time or the resources to catch them all.
Students in Passenger mode are likely to see AI as a means to outsource academic work, dramatically increasing the likelihood of failing to develop “the learning habits necessary to navigate school and work,” write Anderson and Winthrop. Even the 30 percent in Achiever mode—those near the “top of the engagement mountain”—are highly-motivated but often “fragile” learners. For them, the authors say, school can become “all about grades,” making them vulnerable to the allure of outsourcing the occasional task to AI.
This fast-moving new reality makes the work of teaching and assessing students significantly more arduous. Many members of the general public, like journalist Mary Harrington, already recoil at the notion of a smartphone- and social-media-fueled childhood: “On the one hand, a relatively small group of people will retain, and intentionally develop, the capacity for concentration and long-form reasoning,” she writes in The New York Times. “On the other, a larger general population will be effectively post-literate—with all the consequences this implies for cognitive clarity.” The risk of this next generation of AI tools goes “far beyond social media” and is, “scaring the hell out of people,” exclaims social scientist Jonathan Haidt in Politico. “The existential threats are extraordinary. We should be talking about them every day.”
THE HIGH COSTS OF MISUSE
The makers of these paradigm-shifting technologies see the world differently. As new software updates are launched, they imagine—and actively promote—an image of limitless potential. In a 2024 blog post, Google’s head of learning and education Ben Gomes gushed that generative AI was “taking curiosity and understanding to the next level,” and predicted a brave new world: “We’re just at the beginning of how it can help us reimagine learning.”
But in January of this year, Brookings published the results of their premortem analysis of generative AI in K–12 education, a year-long study involving interviews with over 500 teachers, students, families, and education leaders, as well as a review of more than 400 studies. Its conclusion was unambiguous: “We find that at this point in its trajectory, the risks of utilizing generative AI in children’s education overshadow its benefits.”
While the report did not foreclose on possible future uses of AI in K-12 education completely, it cautioned readers against working too hard to find a middle ground. “We worry that the larger education community may miss the forest for the trees,” the authors wrote. “A narrow focus on developing effective AI-supported teaching approaches could obscure how students’ very ability to learn is being undermined by AI overuse, inappropriate use, and non-productive use, both in and, increasingly, outside the classroom.”
Meanwhile, a growing body of research points to serious cognitive and skill-building trade-offs when the software is used.
A recent, high-quality study tracking 27,000 Chinese teens, for example, concluded that 50 percent of students were using AI to complete their homework for them, but when it came time to take tests, AI users’ scores were 20 percent lower than their peers who didn’t use the tech. Homework scores used to be a reliable predictor of performance on a test, a report from The Economist lamented, but now “those who score highest are, perversely, more likely to do worse in exams.”
And in a recent two-year study across 18 different schools, lower-performing middle school students were all provided with access to Khanmigo during the school day—an AI tutor designed to help when asked, but avoid giving up the answers. Ultimately, students “only used Khanmigo about a third of the days they were working in Khan Academy,” reports Matt Barnum for Chalkbeat. When they did interact with the bot, many tried to coax Khanmigo into dishing out the correct answers.
“What this has put in huge spotlight—not that it’s a shock to educators—is how much motivation matters,” Khan Academy’s founder and CEO Sal Khan concluded. Or, as Winthrop and Anderson have mentioned, a lack thereof.
CULTIVATING A DRIVE TO LEARN
Some teachers are taking action to get AI out of their classrooms, returning to familiar standbys like blue books and oral exams.
"There are districts in New Hampshire who have already gone back to turning in assignments not on the Chromebook," explains Megan Tuttle, president of New Hampshire's largest teachers union. "I've been told that the turn-in rate has actually increased." Many have shifted their focus toward ensuring students develop foundational skills by practicing them primarily in class.
Before becoming an assistant principal, former high school English teacher Chanea Bond barred her students from using AI, dedicating their time together to writing and reflecting in composition notebooks. It’s a choice she called one of the “most impactful decisions that I have made as a teacher” a year later: “While some teachers say that AI can serve as a brainstorming partner or an outline generator, I believe it’s so important for students to learn these skills themselves,” she writes. “I’m extremely concerned that reliance on AI will hinder those who need the most support.”
Winthrop says it's up to everyone—from families and policymakers to tech companies, researchers, education systems, and students themselves—to determine how seriously they take the threat of a “great de-skilling” in American schools: “I worry if we let [this] go on too long that we will be in a bad place.” School or district-wide policies restricting AI use require support from students’ parents and caregivers. The work of helping kids develop healthier relationships with generative AI (and technology writ large) can be fortified through consistent conversations and routines that extend outside of the school building and into their homes.
Some schools are foregoing total bans in favor of teaching students responsible AI habits that support their learning, though a lack of guidance and professional development for educators remains a major issue.
In her high school English classroom, Jen Roberts occasionally allows students to enter early drafts into tools like Brisk to get initial feedback on their writing. Middle school science teacher Eric Cross marks each assignment with the level of AI use he considers acceptable and requires students to submit a written disclaimer that details how they used the services. “It’s hard to learn about behavior when we push it into hiding,” he says. Instead of trying to develop AI-proof assessments, Principal Andrew Miller suggests shifting toward thinking about schoolwork on an AI vulnerability spectrum. Educators might ask themselves whether tasks should emphasize completion over thinking: “Don’t ask, ‘Can AI do this?’—instead ask, ‘What thinking is required of students?’”
Still, the question of adolescent motivation and engagement remains, and the equity chasm Winthrop and Anderson predict feels like a possible outcome.
Developmentally, as tweens become teens, students are “no longer motivated by compliance alone,” says former middle school and current high school English teacher Cathleen Beachboard. Instead of asking what do I do next?, they begin asking questions like: who am I? and why does this matter? At this formative stage, purpose is a critically important driver of motivation, explains David Yeager, noted author and psychology professor at the University of Texas at Austin.
“One of the big research insights is that young people really do want to connect to something bigger than themselves,” Yeager says. “Ask them to reflect on their purpose more deliberately, and connect it with what they’re learning in school. And when you do that, adolescents construct their own mental narrative about purpose that helps them to stay engaged even when things are hard or boring.”
That probably won’t be enough. Ultimately, the current crisis of engagement requires an all-hands-on-deck approach. Teachers have always been motivational wizards, but they cannot do the work alone; reciprocity is an overlooked and vital piece of the puzzle.
“I typically talk to students about not just what we are responsible for as teachers, but what the student’s responsibilities are,” says ninth-grade teacher Fatima Belouahi. “Their responsibility is that they engage in the subject matter, even if it’s not something that interests them.” With time, students begin to understand that “the lesson will get better and be more fun if they’re actively participating.”
