The Numbers Behind the Headline
Common Sense Media surveyed 1,017 US teenagers aged 13 to 17 between April 30 and May 14, 2026, with NORC at the University of Chicago running the fieldwork and a margin of error of plus or minus 4.3 percentage points. The topline finding: seven in ten teens now use AI for schoolwork in some form. Among those users, 77 percent describe it as a support tool for brainstorming, checking work, or getting feedback, while 63 percent obtain answers from AI directly, with meaningful overlap between the two groups. This is not a fringe behavior confined to a handful of tech-forward students. It is close to the default mode of doing homework for the majority of American teenagers right now.
The study also breaks down how teens who get direct answers actually use them: 35 percent refine the answer using their own knowledge, 31 percent rewrite it in their own voice, and 25 percent use the AI-generated answer essentially as-is. That last figure is worth sitting with. A quarter of teens who ask AI for an answer are turning it in with minimal alteration, which tells you less about laziness and more about how few of them have been taught what verification or refinement should even look like.
The Instruction Gap Is the Real Story
Adoption without instruction is the pattern that should concern enterprise leaders more than the adoption number itself. Only 27 percent of surveyed teens report that a teacher has ever discussed what AI is or how it works, and only 30 percent have had a classroom conversation about AI safety. That means roughly three in four teens are using a technology daily for academic work that almost no one in their formal education has explained, contextualized, or set boundaries around. The gap between usage and instruction is widening rather than narrowing alongside adoption, because classroom AI curricula take years to design and approve while student usage patterns shift every semester.
This is a policy failure with a workforce consequence that arrives faster than most institutions realize. The teens using AI heavily today without formal instruction are two to four years from internships and entry-level hires, not a full generation away. Whatever gap exists in the classroom right now becomes an onboarding cost that lands directly on employer L&D budgets, whether or not those budgets were built with that specific problem in mind.
What Teens Themselves Think Matters
Perhaps the most useful data point in the study is what teens say they will need for their future, and it is not what most AI-in-education coverage assumes. Working with others ranked highest at 55 percent, followed by critical thinking at 50 percent and complex reading comprehension at 48 percent. Knowing how to use AI effectively ranked far lower, at just 21 percent. Teens are not undervaluing AI fluency because they think it is unimportant, they are undervaluing it because fluency with the tool has become so ambient that it no longer registers as a distinct skill worth naming, the way knowing how to use a search engine stopped registering as a skill a decade ago.
That instinct is not entirely wrong, and it is not entirely right either. Prompting an AI system competently is table stakes now, not a differentiator, which tracks with how little teens prioritize it explicitly. But the skills they do prioritize, critical thinking and collaboration, are precisely the ones needed to catch a confidently wrong AI output before it ships in a report, a customer email, or a piece of code. Teens have correctly identified the skills that will matter. What the data shows is that those skills are not being paired with AI use in the classroom, which is where the actual risk sits.
Two-Thirds Believe AI Helps Understanding, Not Just Speed
A notable finding cuts against the assumption that teens see AI purely as a shortcut. Two-thirds of AI users, 66 percent, believe the tool helps them understand their schoolwork rather than simply finishing it faster, even as 39 percent separately worry they are missing learning opportunities by relying on it. Both things can be true for the same student in the same week, and the data suggests most teens hold that tension consciously rather than ignoring it. This is a more sophisticated relationship with the technology than the usual cheating narrative allows for, and it deserves to inform how employers think about incoming talent rather than assuming universal shortcut-seeking.
The practical takeaway for hiring managers is that this generation is not uniformly AI-dependent in the way that gets described in alarmist coverage. Many of them are actively wrestling with when AI use helps their understanding and when it substitutes for it, without much external guidance on how to draw that line. That self-awareness is an asset an onboarding program can build on, but only if the program acknowledges it exists rather than starting from a blank-slate assumption about what new hires already know.
What This Means for Entry-Level Hiring Assessment
Standard entry-level hiring assessments were built for a world where AI fluency was rare and worth screening for directly. That world is gone. The more useful screen now is not whether a candidate can use AI, nearly all of them can, but whether they can catch an AI-generated error, push back on a confidently wrong output, or explain why they trust or distrust a specific answer. Interview processes and case studies built around producing a polished deliverable increasingly measure prompting skill rather than judgment, and judgment is the scarcer, more valuable trait given what this study shows about the instruction gap these candidates are coming from.
Employers redesigning entry-level assessments this hiring cycle should build in at least one exercise that specifically tests verification behavior: give a candidate an AI-generated output with a subtle but real error and see whether, and how, they catch it. That single exercise will tell a hiring manager more about a candidate's readiness than any question about which AI tools they have used, because tool familiarity is nearly universal now and verification instinct is exactly what this data says is missing.
The Onboarding Investment This Data Justifies
L&D leaders building onboarding programs for the next two to three hiring cycles should treat this data as a planning input, not a curiosity. New hires will arrive with genuine AI fluency and almost no formal grounding in verification, source evaluation, or knowing when to distrust a model's output, because 73 percent of them never had a teacher walk them through even the basics of how the technology works. That gap does not close itself during a standard first-week orientation, and treating it as covered by a generic AI-usage policy slide will leave it exactly where the classroom left it.
A focused, mandatory module on evaluating AI output, distinct from a tool-usage tutorial, is a reasonable and inexpensive addition to any onboarding program serving early-career hires over the next several years. The Common Sense Media data gives L&D leaders a defensible business case for that investment: this is a documented, dated, quantified skills gap, arriving with every graduating class between now and whenever classroom AI literacy instruction finally catches up to classroom AI usage.
