American Students Embrace AI Homework Help Amid Growing Concerns
Between May and December 2025, the proportion of American students—from middle school through college—reporting the use of AI assistance for homework increased significantly, rising from 48% to 62%. This surge in adoption occurred alongside a growing apprehension about the potential consequences of these tools. By December, 67% of surveyed students expressed concern that increased AI use in schoolwork could negatively impact their critical-thinking skills. Despite this, only about one-third of students reported that their schools had established a clear, schoolwide policy governing AI use for homework.
At first glance, these figures might suggest a contradiction: students both fear and rely on AI, while adults offer limited guidance. However, the reality is more nuanced. The term “using AI” encompasses a spectrum of activities ranging from seeking clearer explanations to requesting outright answers. Students themselves differentiate between these uses, and policies frequently vary from one classroom to another.
These insights derive from a RAND Corporation report published in March 2026, which provides a valuable national snapshot of student behavior and attitudes. Importantly, the survey measures perception and self-reported use, but does not establish causation or quantify any decline in reasoning abilities.
A National Snapshot of AI Use in Education
RAND’s December 2025 survey encompassed 1,214 enrolled Americans aged 12 to 29, including 184 middle school students (grades 5–8), 738 high school students (grades 9–12), and 292 college undergraduates. Participants represented a cross-section of educational settings, including public, charter, private, homeschool, and postsecondary institutions. The responses were weighted to reflect the demographic composition of enrolled US youth by gender, race or ethnicity, grade level, and geographic region.
It is important to note that the May and December surveys were independent cross-sectional snapshots rather than longitudinal tracking of the same individuals. The May survey included 2,039 respondents aged 12 to 21, while the December survey expanded eligibility to ages 12 to 29 and altered college sampling methods. Consequently, while the increase from 48% to 62% in AI use is meaningful, researchers caution against overinterpreting trends among college students or conducting strict statistical comparisons across waves.
Middle and High School Propel the Growth
The rise in AI homework use was particularly pronounced among younger students. Middle schoolers increased their reported use from 30% in May to 46% in December, while high school students rose from 49% to 63%. College student use remained high but relatively stable (71% to 72%), though sampling changes advise caution in interpreting this figure.
December’s survey further broke down AI tool categories, finding that 71% of respondents used at least one type. Chatbots were the most popular at 60%, followed by AI writing assistants (21%), general homework helpers (15%), and AI tutors (8%). ChatGPT led as the dominant named product, with 53% reporting its use, up from 45% in May. Google Gemini usage more than doubled from 13% to 28%. These figures reflect usage occurrence, not intensity or whether AI-generated content was submitted as original work.
This widespread adoption aligns with patterns observed in other domains, where AI assistance helps less-experienced individuals complete tasks more efficiently. However, education differs fundamentally from task completion centers: the learning process emphasizes mental effort and skill development, which can be undermined by overreliance on AI-generated answers.
AI Use Encompasses Diverse Purposes
When asked why they used AI for schoolwork, students gave a variety of reasons without a single dominant purpose. Thirty-eight percent sought clearer explanations of assignments, 35% used AI for brainstorming ideas, while 33% each reported looking up facts and drafting or revising written work.
Most students selected two of the eight purposes offered, and open-ended responses revealed nuanced uses such as checking answers, understanding mistakes, or requesting step-by-step guidance through problems. These activities differ significantly in educational value. For example, asking “Why does this equation require that step?” fosters understanding, whereas copying an AI-provided solution does not.
This diversity in use explains why students might frequently engage AI without uniformly viewing all uses as acceptable or legitimate.
Students Define Their Own Boundaries Around Cheating
Students’ perceptions of what constitutes cheating varied by AI use case. Among those using AI to obtain better explanations, 78% did not consider it cheating. For brainstorming, 72% disagreed that it was cheating, and 67% felt the same about fact-checking. However, when AI was used to generate homework answers directly, only 45% rejected the practice as cheating, with many others expressing situational ambivalence.
It is important to note that only students who admitted to a particular AI use were asked about cheating in that context, potentially underestimating the extent to which some uses are viewed as improper.
Overall, students tend to accept AI as a tool for clarification, ideation, and background research, but remain divided over its use for producing final, submitted work.
Concern Over Critical Thinking Rises, but Use Grows
The survey question gauging concern about AI’s impact on critical thinking asked students to agree or disagree with the statement: “The more students use AI (such as ChatGPT) for their schoolwork, the more it will harm their critical thinking skills.” Agreement increased from 54% in February 2025 to 67% in December.
Concern was evident across all educational levels, rising most sharply among middle school students. Notably, even among AI users, 60% believed that heavier AI use could harm critical thinking, compared to 78% among non-users. Differences across educational stages within the user group were not statistically significant.
These findings reflect perceptions rather than measured cognitive outcomes. The RAND report did not include objective assessments of reasoning skills or experimental designs to determine causality. Nonetheless, students’ concerns may indicate an awareness of changes in their engagement or effort.
Reconciling Use and Concern: Not Necessarily Hypocrisy
Students can simultaneously recognize the risks of overreliance on AI while appreciating its targeted benefits. Just as calculators save time without replacing all arithmetic practice, AI can efficiently support learning when used judiciously.
Previous research into human-AI creativity comparisons highlights a similar distinction: AI-generated output quality does not equate to demonstrated learning or skill acquisition. Thus, the coexistence of widespread AI use and rising concern likely reflects an ongoing negotiation of this trade-off.
AI offers speed, accessibility, and sometimes valuable explanations, but it can also enable shortcuts that bypass essential cognitive effort.
Many Schools Lack a Unified AI Policy
Only about one-third of students across middle school, high school, and college reported that their schools had a clear, schoolwide rule regarding AI use for homework. Where policies existed, prohibitions outnumbered allowances by two to three times.
For many students, AI policy depended on individual teachers: approximately 34% of middle schoolers, 44% of high schoolers, and 57% of college students reported this variability. Others indicated no rules or were unsure. Importantly, these data reflect students’ understanding of policies rather than formal audits of written rules.
Policies hidden in handbooks may be ineffective if students are unaware, whereas assignment-specific teacher instructions can provide clearer guidance in the absence of a schoolwide framework.
Teacher discretion can be beneficial when balancing appropriateness across different assignments, but inconsistency risks unfair penalties if students are unaware of expectations.
Earlier RAND research on AI guidance in schools similarly found that AI use outpaced clear institutional direction. The December 2025 findings underscore this gap from the student perspective: teachers may make local decisions, but many learners lack a consistent baseline for AI use.
Effective AI Policies Focus on Who Does the Thinking
RAND recommends moving beyond a binary “AI allowed versus not allowed” framework toward distinguishing between cognitive offloading and cognitive augmentation. Cognitive offloading occurs when AI performs the mental work for the student (e.g., generating a final answer), while cognitive augmentation involves AI prompting the student to engage more deeply or carefully with material (e.g., offering counterarguments or feedback).
This distinction, while not always clear-cut, can help shape actionable policies. For example, submitting an AI-generated answer without modification is offloading, whereas requesting AI to critique one’s reasoning preserves independent thought.
Effective guidance should operate on two levels: a schoolwide baseline addressing disclosure, privacy, authorship, and unacceptable substitution; and teacher-level instructions clarifying which AI uses are appropriate for specific assignments and what evidence of independent thinking is required.
Context matters. AI feedback on a draft can facilitate revision and learning, but generating an initial draft before forming one’s own argument may undermine the intended learning process.
Policies must also be communicated in clear, practical language students can apply independently. Vague admonitions like “use AI responsibly” are insufficient, whereas specific guidelines such as “You may ask for explanations and practice questions but may not ask AI to write your response; include your prompts” provide actionable direction.
Schools Must Act Now to Shape AI’s Role in Learning
The RAND report advises against panic but warns that waiting for perfect evidence while students independently navigate AI use is untenable. By late 2025, AI assistance with homework was already normalized for the majority of US students surveyed.
Many students simultaneously express unease about AI’s impact on critical thinking and distinguish between legitimate and illegitimate uses. This suggests students possess a developing, if informal, understanding of AI’s educational role even as formal policies lag behind.
The most constructive response is neither a blanket ban nor a vague endorsement. Instead, schools should clearly articulate the learning objectives of assignments, identify which forms of AI assistance preserve those goals, and explicitly forbid those that undermine them.
AI has already entered students’ homework routines. The critical question is whether schools will meet this reality with clear, consistent, and intelligible expectations or leave students to navigate a patchwork of assumptions, habits, and penalties. The rapid 62% adoption rate highlights the pace of change, while the 67% concern rate underscores that students recognize what is at stake.
For more on this evolving topic, see the full article Here.
