When AI does the homework: why good grades no longer mean learning happened

By Zendoric · September 11, 2026.

Dr. Khilia Giacobone, director of children and families at Youth for Tomorrow —a Virginia (U.S.) organization that works with families— says artificial intelligence is an increasingly frequent topic of conversation with the parents she advises. Her diagnosis, reported by local television station WWBT: students can use AI to get good grades, but that does not mean they are learning the material.

Giacobone sums it up with an everyday example: asking AI to solve a problem is efficient in the moment, but failing to master the basics takes its toll later on. Her prescription is not to ban the tool, but for parents to get involved: talk to their children about their schoolwork and about the learning process, not just about whether the assignment is "done," and take advantage of everyday situations —she gives the example of going shopping and comparing prices— to practice arithmetic and planning without shortcuts. She also calls for building tolerance for frustration: not everything should be easy or immediate.

There is a second warning, more troubling than the academic one: blind trust. Giacobone recounts asking someone she advises what happened when AI gave a wrong answer, and the reply was that this "doesn't happen." Her response: it does happen, and what's more, the system returns generic advice that looks tailored without being so. It is a nuance many adults —not just children— have yet to internalize, and one Giacobone translates into a simple recommendation: parents still have to keep an eye on what goes on at home, with or without AI in the mix.

As sector context, this scene is repeating itself in thousands of homes and classrooms since conversational tools arrived en masse in homework assignments: teachers detecting work done entirely by AI, families with no clear criteria on when to allow it, and an education policy vacuum that each school —or each parent— resolves as best it can. Pedagogy has spent decades documenting the value of "productive struggle": the effort of getting it wrong, persisting and arriving at the answer on your own as the engine of deep learning. Used as a shortcut, AI is precisely what neutralizes it.

The education industry has already picked up on this: the products being pushed hardest right now —from AI tutors built into school platforms to assistants designed to "guide without giving the answer"— compete on that exact nuance, not on raw model power. Differentiation is no longer about answering better, but about resisting the temptation to answer directly and instead asking the questions that push the student to think. It is a still-emerging product shift, and much of whether AI in the classroom ends up being a prosthesis or a coach depends on it.

Our reading: what Giacobone describes is not a quirk of one particular family, it is the first visible symptom of a tension society will have to resolve on a larger scale. In the short term, the gap will not be set by who has access to AI —that is already almost universal— but by who has an adult nearby capable of turning it into a copilot rather than a substitute. That widens, rather than narrows, educational inequality between households with different amounts of time and attention available to supervise. It is the same dynamic we have already seen in the classroom, where the teacher who orchestrates AI as a teaching tool wins and the one who lets it replace them loses; at home, the pattern repeats with parents.

In the long run, however, the same technology that today makes it possible to cheat without learning is the one that, well designed, can bring every child something close to what education research has spent decades identifying as the most effective method there is: a personal tutor who adapts to their pace, available at all times, without the cost that today only a few families can afford. That is the ground on which the future of AI in education is really being decided: not whether children are going to use it, which they already do, but whether we manage to design it —and teach parents to use it— in a way that trains thinking instead of replacing it.