Artificial intelligence is not the enemy. I use it constantly. For adults who already know how to think, write, research and solve problems, AI can be a genuine force multiplier - saving time, sharpening structure and speeding up routine work. But that is not the same as education. The problem is not whether AI is useful. It is. The problem is what happens when students lean on it before they have built the basic skills it is supposed to support. A professional using AI to accelerate work they already understand is one thing. A high school or college student using it to replace reading, thinking, outlining and writing is something else entirely.

That is where the concern gets serious. It is a concern for all young people still learning. But it may be especially dangerous for groups that have historically struggled more with graduation rates, academic achievement, chronic absenteeism and other barriers tied to family and environment. The issue is not race in isolation. The issue is whether AI becomes a shortcut layered on top of already shaky foundations.

AI Can Help - But Only When It Supports Learning Rather Than Replacing It

Used correctly, AI can be a good educational tool. It can explain concepts in plain language, generate practice questions, give instant feedback, help multilingual learners and offer individualized support that many students do not get consistently in overcrowded classrooms. OECD has noted that generative AI can improve learning and foster critical thinking, creativity and collaboration when used with clear pedagogical intent. UNESCO has argued similarly that AI can help address major educational challenges when implemented with human-centered safeguards.

That is the best-case version. The worst-case version is different. In that version AI does not support learning - it replaces it. The student no longer wrestles with the reading, no longer organizes the argument, no longer chooses evidence, no longer revises language and no longer develops the mental endurance that real education requires. Brookings recently warned that when children replace effortful learning with generative AI to shortcut assignments, it harms cognitive development. Research from Microsoft has linked higher AI use and cognitive offloading to weaker critical thinking. That is the line that matters. AI as tutor can help. AI as substitute hollows students out.

A Diploma Is Not the Same as Mastery

One reason this issue matters so much is that schools already struggle with the gap between credentialing and competence. The national public high school adjusted cohort graduation rate was 87 percent in 2021-22, with rates of 90 percent for White students, 83 percent for Hispanic students and 81 percent for Black students. Those are not collapse-level numbers but they show meaningful gaps. And graduation rates do not tell the whole story. NAEP continues to document persistent achievement gaps between Black and White students and between Hispanic and White students in reading and math. NCES describes these as statistically significant and has tracked them over time. In other words, a diploma can coexist with weak academic readiness - and has been doing so for years.

That is why AI introduces a new problem. If schools are already passing too many students who have not fully mastered the material, AI makes that easier. It produces cleaner prose, better formatting and more polished answers than the student could generate alone - which means students may look more competent on paper than they actually are. The danger is not just cheating. The danger is social promotion with better grammar.

Why This Could Hit Some Groups Harder

This is where the conversation gets uncomfortable but needs to stay honest. Recent Pew research found that Black and Hispanic teens are more likely than White teens to report frequent chatbot use for schoolwork and more likely to say they do all or most of their schoolwork with chatbot help. About six in ten Black or Hispanic teens reported using chatbots for schoolwork, compared with roughly half of White teens. Earlier Pew reporting found daily chatbot use among Black and Hispanic teens running above the level reported by White teens.

That does not prove harm by itself. It does raise a serious possibility. If a population is already more likely to face academic headwinds, and also more likely to rely heavily on AI for schoolwork, the technology may end up reinforcing weakness instead of curing it. The point is not that Black or Hispanic students are less capable. The point is that students who begin from a weaker position are more vulnerable to using AI as a crutch instead of a ladder. That is true for poor White students too. It is true for anyone whose school preparation is thin, whose home environment is chaotic, whose attendance is weak or whose reading and writing skills are already behind. The risk is not racial destiny. The risk is compounded fragility.

Dependency Before Competence Is the Core Problem

Adults who earned real competence before AI generally know when the machine is helping and when it is bluffing. Teenagers and undergraduates often do not. That is the heart of the issue. A student who has never built the muscles of concentration, revision and critical judgment is likely to offload those tasks before those muscles even form. That is not empowerment. That is dependency dressed as efficiency.

Brookings has argued that the future of AI in education depends on whether systems are designed to help students prosper, prepare and protect themselves in an AI world. OECD has made a similar point, stressing that effective use depends on pedagogy, safeguards and human judgment. Those are not abstract concerns. They go directly to whether AI builds capability or erodes it. If schools do not draw that line clearly, the likely result is predictable: the strongest students will use AI as leverage while the weakest students will use it as concealment. That is how inequality widens under the banner of innovation.

This Is Not an Argument for Banning AI

Trying to ban AI outright would be foolish. Students are going to use it. Professionals already do. Employers will expect familiarity with it. The goal should not be to pretend the technology does not exist. The goal should be to force visible thinking.

That means assignments that show process, not just polished output. More in-class writing. More oral defense. More staged drafts. More source verification. More reflection on why students made certain choices. If a student submits an elegant paper but cannot explain the argument, defend the evidence or revise the logic without machine help, then the learning did not happen. AI should be permitted as a tool but not as a replacement for cognition. That distinction matters for everybody - affluent students, poor students, students of every background. But it matters especially for young people whose educational footing is already unstable, because they have more to lose from fake competence than anyone else.

My Bottom Line

AI is not automatically making students smarter and it is not automatically making them dumber. It depends entirely on how they use it and when. For learners who already have strong foundations, AI can be a genuine asset. For learners who are still building those foundations - especially in communities already dealing with lower academic readiness, chronic absenteeism, weaker family support and uneven school quality - AI can become a dangerous illusion. It can help them pass without helping them grow.

A society that hands out credentials without competence is setting young people up for failure. The short-term result is more passing, more graduating and better-looking work on paper. The long-term result is adults who have been told they are ready when they are not. That is bad for employers, bad for colleges, bad for civic life and worst of all bad for the very students the system claims it is helping.

If AI becomes a mask for weak learning instead of a support for real learning, it will not close opportunity gaps. It will hide them until reality shows up.