Should Schools Adopt AI Before the Evidence Is In?
September 9, 2026 by
Artificial intelligence is already in American schools, whether or not schools have decided what to do about it. A Common Sense Media census released in June 2026 found that 86% of children aged 9-17 use or interact with generative AI, with nearly a quarter using it daily, and among young AI users, 85% use it for schoolwork.1 A 2025 RAND survey found that 54% of students reported using AI for school—an increase of more than 15 percentage points over the prior 1-2 years—and that 53% of English language arts, math, and science teachers used AI for their work.2
States are beginning to draft legislation to address AI usage in schools. Virginia’s new statute directing its education department to issue guidance on AI in schools took effect July 1, and legislatures in more than 30 states have introduced AI-in-education bills this year.3
Behind the adoption race sits an uncomfortable fact. In March 2026, Stanford University’s SCALE Initiative published a comprehensive review of the research on AI in K-12 education and reached a striking conclusion: Out of more than 800 academic papers examined, only 20 met the bar for high-quality causal evidence—and none of those studied student AI use in U.S. K-12 classrooms. Schools, in other words, are making decisions about AI ahead of the evidence. The question this post examines is whether that is a reason to slow down.
What Does the Research Show?
The Stanford review, conducted by researchers Lily Fesler, JP Martinez Claeys, Chris Agnew, and Susanna Loeb, sheds important light on the limits of our current knowledge. The 20 rigorous studies it identified offer some encouraging signals: AI appears most effective when students use it for math practice, writing tasks, and programming, and it works better when it offers hints after wrong answers rather than simply supplying correct ones. For teachers, early studies suggest AI can reduce time spent on lesson preparation while maintaining instructional quality.4
But the review is equally notable for what the research cannot yet say. Most studies examine short-term task performance—sometimes a single 20-minute session—rather than durable learning. Much of the work involves students older than 18 or was conducted outside the United States. And very little research examines effects on equity, student wellness, or social development.5 The authors note that because tools change faster than studies can be published, a product rigorously evaluated today may be a different product by the time the findings appear.6
Survey data fill in the picture of attitudes, if not effects. A Gallup-Walton Family Foundation study found that about 60% of teachers used AI during the 2024-25 school year, and those using it at least weekly saved an average of 5.9 hours per week—the equivalent of six weeks per school year.7 Yet, the same teachers were wary about students: 57% said weekly AI use would decrease students’ independent thinking, and 52% said it would erode critical thinking.8 Parents are warier still. In the RAND survey, 61% of parents agreed that greater AI use will harm students’ critical-thinking skills—nearly three times the 22% of school district leaders who said the same.9
What Is Being Proposed?
State legislatures in 2026 have largely avoided either mandating or banning classroom AI. Instead, according to FutureEd’s legislative tracker, this year’s bills concentrate on three themes: protecting student data privacy, requiring human oversight of consequential decisions, and securing parental consent or opt-out rights. Oklahoma’s newly enacted Responsible Technology in Schools Act requires every district to adopt a written AI policy; prohibits AI from serving as the primary basis for grading, discipline, or placement; and preserves parents’ right to opt students out of student-facing AI tools.10 Virginia’s S.B. 394, which passed the House of Delegates 95-0, directs the state education department to issue guidance addressing student data privacy and teacher training, and creates a pilot program to fund and evaluate AI platforms for instruction and tutoring.11
Arguments in Favor of Bringing AI Into Classrooms
Proponents see in AI a chance to deliver what education technology has long promised and rarely achieved: individualized attention for every student. John Bailey, a senior fellow at the conservative-leaning American Enterprise Institute, has argued that AI tutors can consistently adapt to each student’s needs and, with their surprising emotional intelligence, may motivate and reassure students in ways that stretch the capacity of even skilled teachers. In testimony before Congress, he urged urgency: “We need more experimentation with AI tutors to personalize instruction and close gaps.”12 With student achievement still depressed from pandemic-era losses, he argued, delaying a tool that could provide individualized feedback at scale is itself a costly choice.
A second argument concerns the world students are entering. Michael Horn, an author and lecturer at the Harvard Graduate School of Education, has argued that because the workforce is already incorporating AI, schools that avoid it leave students unprepared—and he has criticized educators who are “more focused on cheating and stopping AI usage than on how they can use AI to make education better.”13 Notably, Horn’s endorsement is conditional rather than sweeping. Asked whether schools should adopt AI, he answered, “It depends,” adding: “Would I tell a school not to use something like Amira Learning, an AI-powered app for building literacy that has good efficacy studies behind it? Not at all. It could make a lot of sense if it’s part of a coherent instructional model.”14
Supporters also point to what AI is already doing for teachers. Beyond the time savings documented by Gallup, majorities of AI-using teachers have reported that the tools improve the quality of their work, and nearly 60% have said that AI improves the accessibility of learning materials for students with disabilities. As Gallup Research Director Zach Hrynowski put it, “It’s never going to mean that students are always going to be taught by artificial intelligence and teachers are going to take a backseat.”15
Arguments Against Rushing AI Into Classrooms
Skeptics start with the evidence gap itself. If there are no high-quality studies of student AI use in U.S. K-12 classrooms, they argue, then schools adopting student-facing AI are running an uncontrolled experiment on children—one whose effects on long-term learning, equity, and development no one has measured. Andy Smarick of the Manhattan Institute has taken this argument furthest, calling for a two-year moratorium on AI in public education to prevent a rushed and risky rollout.16
The same Horn who defends evidence-backed tools has also issued a warning that cuts the other way: “Mindlessly layering ‘AI’ for its own sake into a school or classroom doesn’t make sense. And it could very well be detrimental,” he has said, adding that he worries “a lot about the lack of coherence in schools,” a problem he sees getting worse.17 On this view, the risk is not AI itself but a repeat of past ed-tech cycles in which schools bought technology first and asked what it was for later.
Some educators argue that the right response to AI is better teaching, not more AI. Carlo Rotella, an English professor at Boston College, wrote in the New York Times Magazine that the threat of AI-enabled cheating has pushed him toward practices that are both effective and largely AI-proof: handwritten blue-book exams, frequent low-stakes quizzes, requiring students to show their work, and grading class participation.18 His experience suggests that traditional academics can survive AI—but it also implies that student learning may depend on keeping AI out of the loop for core intellectual work, not building it in.
Students and parents supply the bluntest skepticism. Ashanty Rosario, a senior at a New York public high school, wrote in The Atlantic under the headline “I’m a High Schooler. AI Is Demolishing My Education” that she watched a classmate paste an assigned Frederick Douglass chapter into ChatGPT and turn in the resulting annotations for participation credit. “These programs have destroyed much of what tied us together as students,” she argued.19 Students’ own doubts appear to be deepening as their use grows. In RAND’s youth panel, the share using AI for homework rose from 48% in May 2025 to 62% in December, while the share saying AI use for schoolwork harms critical thinking rose from 54% to 67% over the same period.20 To critics, the earlier RAND finding that district leaders are far less worried than parents suggests the people making adoption decisions are out of step with the families affected by them.21 Early markers of dependence give that worry some grounding: Among children who use AI for homework, 16% told Common Sense Media they have trouble starting or finishing assignments without it, rising to 28% among daily homework users.22
Broader Questions
There is a paradox running beneath this debate, and it has not gone unnoticed. At the same moment that states are restricting phones to get screens out of students’ hands, many are encouraging schools to embrace AI—a tension that education scholars Horn and Rick Hess examined at length in an exchange this spring titled “Schools Are Urged to Embrace AI—and Ban Phones. Can We Resolve the Tension?” Horn’s answer is that both impulses can be coherent if technology serves a deliberate instructional design rather than being layered on by default.23 Whether schools can manage that discipline is an open question. A 2026 Gallup study found that fewer than one in ten teachers receive formal guidance on how to use AI for any specific work task.24 The evidence will eventually catch up to the adoption. The question is what schools will have already decided by then.
Discussion Questions
- How are you and your classmates actually using AI for schoolwork, if at all? Do your school’s rules—written or unwritten—match what really happens?
- Stanford researchers found no high-quality causal studies of student AI use in U.S. K-12 classrooms. Should schools wait for rigorous evidence before adopting new technology, or does waiting carry its own risks? What would you want to know before deciding?
- John Bailey argues AI tutors can give every student individualized attention, while Carlo Rotella responds to AI by using blue books, quizzes, and participation—practices that keep AI out. Which vision of the classroom do you find more persuasive, and why?
- Only 22% of district leaders believe AI will harm students’ critical thinking, compared with 61% of parents. What might explain that gap, and whose judgment should carry more weight in adoption decisions?
- Most 2026 state legislation focuses on privacy, human oversight, and parental consent rather than mandating or banning classroom AI. Is that the right role for lawmakers? What, if anything, is missing?
- If you served on a school board considering an AI tutoring program, what evidence would you require before approving it? What would make you vote no?
Related Posts
As always, we encourage you to join the discussion with your comments or questions below.
Close Up is proud to be the nation’s leading nonprofit civic education organization, working with schools and districts across the country since 1971. If you would like to partner with us or learn more about our experiential learning programs, professional development, or curriculum design and consulting, contact us today!
Sources
Image Credit: krungchingpixs | Shutterstock
[1] Common Sense Media, “The Common Sense Media Census: AI Use by Tweens and Teens,” June 8, 2026: https://www.commonsensemedia.org/sites/default/files/research/report/2026-ai-use-by-tweens-and-teens-1.pdf; press release: https://www.commonsensemedia.org/press-releases/common-sense-media-releases-inaugural-annual-study-on-ai-use-by-tweens-and-teens
[2] Christopher Joseph Doss, Robert Bozick, Heather L. Schwartz, Lisa Chu, Lydia R. Rainey, Ashley Woo, Justin Reich, and Jesse Dukes, “AI Use in Schools Is Quickly Increasing but Guidance Lags Behind: Findings from the RAND Survey Panels,” RAND Corporation, RR-A4180-1, September 2025: https://www.rand.org/pubs/research_reports/RRA4180-1.html
[3] FutureEd, “Legislative Tracker: 2026 State AI in Education Bills,” updated July 13, 2026: https://www.future-ed.org/legislative-tracker-2026-state-ai-in-education-bills/
[4] Lily Fesler, JP Martinez Claeys, Chris Agnew, and Susanna Loeb, “The Evidence Base on AI in K-12: A 2026 Review,” AI Hub for Education, SCALE Initiative, Stanford University, 2026: https://scale.stanford.edu/sites/default/files/The%20Evidence%20Base%20on%20AI%20in%20K-12%20Report.pdf
[5] Stanford SCALE Initiative, “Understanding the Evidence Base on AI in K-12 Education”: https://scale.stanford.edu/research-in-action/understanding-evidence-base-ai-k12-education
[6] Lily Fesler, JP Martinez Claeys, Chris Agnew, and Susanna Loeb, “The Evidence Base on AI in K-12: A 2026 Review,” AI Hub for Education, SCALE Initiative, Stanford University, 2026: https://scale.stanford.edu/sites/default/files/The%20Evidence%20Base%20on%20AI%20in%20K-12%20Report.pdf
[7] Walton Family Foundation, “The AI Dividend: New Survey Shows AI Is Helping Teachers Reclaim Valuable Time,” June 25, 2025: https://www.waltonfamilyfoundation.org/the-ai-dividend-new-survey-shows-ai-is-helping-teachers-reclaim-valuable-time; Gallup, “Three in 10 Teachers Use AI Weekly, Saving Six Weeks a Year”: https://news.gallup.com/poll/691967/three-teachers-weekly-saving-six-weeks-year.aspx
[8] The 74, “Survey: 60% of Teachers Used AI This Year and Saved up to 6 Hours of Work a Week,” July 8, 2025: https://www.the74million.org/article/survey-60-of-teachers-used-ai-this-year-and-saved-up-to-6-hours-of-work-a-week/
[9] Christopher Joseph Doss, Robert Bozick, Heather L. Schwartz, Lisa Chu, Lydia R. Rainey, Ashley Woo, Justin Reich, and Jesse Dukes, “AI Use in Schools Is Quickly Increasing but Guidance Lags Behind: Findings from the RAND Survey Panels,” RAND Corporation, RR-A4180-1, September 2025: https://www.rand.org/pubs/research_reports/RRA4180-1.html
[10] FutureEd, “Legislative Tracker: 2026 State AI in Education Bills,” updated July 13, 2026: https://www.future-ed.org/legislative-tracker-2026-state-ai-in-education-bills/
[11] Williams Mullen, “Virginia To Establish Guidance for Artificial Intelligence in Schools,” July 2026: https://www.williamsmullen.com/insights/news/legal-news/virginia-establish-guidance-artificial-intelligence-schools; VPM News, “Spanberger to consider legislation requiring state guidance on AI in schools,” March 25, 2026: https://www.vpm.org/generalassembly/2026-03-25/hb1186-sb394-ai-guidance-schools-pekarsky-rasoul-spanberger-heny
[12] John Bailey, “AI and the Future of Learning,” testimony, American Enterprise Institute, October 2025: https://www.aei.org/research-products/testimony/ai-and-the-future-of-learning/; John Bailey, “AI Tutors: Hype or Hope for Education?” Education Next: https://www.educationnext.org/ai-tutors-hype-or-hope-for-education-forum/
[13] Nat Malkus, “AI Has Come for K–12 Education” (summarizing Michael Horn in Education Next): https://www.aei.org/education/ai-has-come-for-k-12-education/
[14] Rick Hess and Michael Horn, “AI in K–12: A New Year Reality Check for School Leaders,” Education Week Opinion, January 2026: https://www.edweek.org/technology/opinion-ai-in-k-12-a-new-year-reality-check-for-school-leaders/2026/01; also published as “How Are K–12 School Leaders Managing the Use of AI?” Education Next: https://www.educationnext.org/how-are-k-12-school-leaders-managing-the-use-of-ai/
[15] The 74, “Survey: 60% of Teachers Used AI This Year and Saved up to 6 Hours of Work a Week,” July 8, 2025: https://www.the74million.org/article/survey-60-of-teachers-used-ai-this-year-and-saved-up-to-6-hours-of-work-a-week/
[16] Andy Smarick, “The Case for an AI Moratorium in Schools,” The Dispatch: https://thedispatch.com/article/artificial-intelligence-education-learning-moratorium/
[17] Rick Hess and Michael Horn, “AI in K–12: A New Year Reality Check for School Leaders,” Education Week Opinion, January 2026: https://www.edweek.org/technology/opinion-ai-in-k-12-a-new-year-reality-check-for-school-leaders/2026/01; also published as “How Are K–12 School Leaders Managing the Use of AI?” Education Next: https://www.educationnext.org/how-are-k-12-school-leaders-managing-the-use-of-ai/
[18] Carlo Rotella, “I’m a Professor. A.I. Has Changed My Classroom, but Not for the Worse,” New York Times Magazine, November 25, 2025: https://www.nytimes.com/2025/11/25/magazine/ai-higher-education-students-teachers.html
[19] Ashanty Rosario, “I’m a High Schooler. AI Is Demolishing My Education,” The Atlantic, September 3, 2025: https://www.theatlantic.com/technology/archive/2025/09/high-school-student-ai-education/684088/
[20] Heather L. Schwartz and Melissa Kay Diliberti, “More Students Use AI for Homework, and More Believe It Harms Critical Thinking: Selected Findings from the American Youth Panel,” RAND Corporation, RR-A4742-1, March 17, 2026: https://www.rand.org/pubs/research_reports/RRA4742-1.html
[21] Christopher Joseph Doss, Robert Bozick, Heather L. Schwartz, Lisa Chu, Lydia R. Rainey, Ashley Woo, Justin Reich, and Jesse Dukes, “AI Use in Schools Is Quickly Increasing but Guidance Lags Behind: Findings from the RAND Survey Panels,” RAND Corporation, RR-A4180-1, September 2025: https://www.rand.org/pubs/research_reports/RRA4180-1.html
[22] Common Sense Media, “The Common Sense Media Census: AI Use by Tweens and Teens,” June 8, 2026: https://www.commonsensemedia.org/sites/default/files/research/report/2026-ai-use-by-tweens-and-teens-1.pdf; press release: https://www.commonsensemedia.org/press-releases/common-sense-media-releases-inaugural-annual-study-on-ai-use-by-tweens-and-teens
[23] Rick Hess and Michael Horn, “Schools Are Urged to Embrace AI—and Ban Phones. Can We Resolve the Tension?” Education Week Opinion, April 2026: https://www.edweek.org/technology/opinion-schools-are-urged-to-embrace-ai-and-ban-phones-can-we-resolve-the-tension/2026/04
[24] Gallup, “Most Teachers Receive No Formal Guidance on AI Use,” May 2026: https://news.gallup.com/poll/710534/teachers-receive-no-formal-guidance.aspx
