For years, there has been a quiet understanding in many high schools that success in certain courses, especially senior math and sciences, required something extra. Not more effort or better attendance, but a tutor. Parents would trade recommendations, students would quietly admit they needed one, and tutoring centres would advertise that “everyone needs help.” In some, especially affluent communities, paid tutors became part of the culture, almost an unspoken prerequisite to keeping up.
That world may be coming to an end.
AI has entered the tutoring business, and it does not take nights or weekends off. For the first time, students have access to personalized, immediate feedback and explanations any time they need it. They can ask follow up questions without embarrassment, get alternative explanations and have complex problems broken into smaller steps. All of this is available for free, or for the price of a phone app. The model that tutoring companies built around scarcity and exclusivity is being replaced by abundance and accessibility.
It is not only about convenience. Tools like ChatGPT, Claude, and Magic School AI can act as math coaches, writing mentors and language partners. They remember the work, adapt to a student’s level, and adjust explanations when the learner gets stuck. The value proposition that human tutors once held, personalization, is becoming a default feature of modern AI systems.
Just last week, one of our Grade 12 students shared how she had been struggling with integration by parts in calculus. Instead of waiting for a weekly tutoring session, she worked through problems with an AI tutor at 11 p.m., asking it to explain the same concept three different ways until it clicked. “It never got frustrated when I asked the same question again,” she said. “And I could be honest about what I did not understand.”
When I first started drafting this piece, I was ready to declare the end of the tutoring era. The evidence seemed clear. The assumption that you need a tutor to survive Pre Calculus is being upended. For many students, the AI sitting quietly on their laptop or phone now fills that role, often better and more patiently than the Saturday morning sessions they once dreaded.
Then I started reading the research. And my thinking got more complicated.
What the Research Actually Shows
The October 2025 edition of AASA’s School Administrator magazine dedicates significant space to the state of tutoring in American schools. AASA is an American based organization, but the questions it raises cross borders easily. The tension between equitable access and quality instruction, the challenge of sustaining initiatives beyond initial funding, the promise and limits of technology in supporting learners: these are Canadian conversations too. The research may come from Texas and Massachusetts, but it speaks directly to what we are wrestling with in British Columbia and across the country.
Liz Cohen, in her article drawing from her book The Future of Tutoring: Lessons from 10,000 School District Tutoring Initiatives, documents an unprecedented expansion. Within a year of the pandemic’s onset, 10,000 U.S. school districts were offering some form of tutoring after years of almost none. By May 2024, 46 percent of public schools reported providing high dosage tutoring, and just 13 percent said they offered no tutoring at all.
Research from the Johns Hopkins Center for Research and Reform in Education, featured in this issue, offers evidence that virtual tutoring with human tutors can produce meaningful results. Grade one students assigned to Air Reading, a structured virtual tutoring program, four times a week for a semester gained nearly 1.6 additional months of learning. Those who attended at least 40 sessions saw even greater progress.
But here is the tension that caught my attention: the research consistently shows that the most effective tutoring models still rely on human tutors. Studies on AI tutoring directly with students remain in early stages, and even the most promising work positions AI as supporting human tutors rather than replacing them
I had to sit with that for a while.
The Hybrid That Works
One case study which helped my framing was learning about the work happening in Ector County ISD in Texas. In partnership with Stanford University, they developed something called Tutor CoPilot. It uses AI not to tutor students directly, but to coach human tutors in real time, suggesting questions to ask, concepts to revisit, hints to offer.
The results are striking: students whose tutors used the AI prompts scored 14 percentage points higher than those whose tutors did not. The AI shifted tutors toward stronger pedagogy, guiding student thinking rather than simply giving away answers. And here is the part that matters most for equity: the greatest benefits went to less experienced tutors. The tool essentially democratized tutoring quality, helping novice tutors perform nearly as well as veterans.
This is not AI replacing humans. This is AI and humans amplifying each other.
What AI Cannot Yet Do
Cohen’s research surfaces something that pure AI cannot yet replicate. The success of tutoring, she argues, is deeply rooted in human relationships. It helps young people feel they matter. It builds motivation through productive struggle in a high support, high standards environment Cohen (This podcast is also a good background on Cohen’s work).
There will still be families who seek human tutors, especially for accountability or emotional connection. Some students need the structure of showing up, the social pressure of not wanting to disappoint someone, or simply the reassurance of a person saying “you’ve got this.” AI has not yet mastered the art of knowing when a student needs a break, a pep talk, or someone to believe in them.
The question is whether it will, and how soon.
The New Digital Divide
For schools, this raises urgent questions. Do we teach students how to use AI tutors effectively? How do we ensure that all students, not only the digitally confident, benefit from these new tools?
The digital divide is no longer just about device access. It is also about knowing how to prompt effectively, when to question an AI response, and how to use these tools for learning rather than answer getting. A student with strong digital literacy might turn ChatGPT into a Socratic tutor. Another might never get past using it as a homework completion machine. If we are not careful, digital confidence becomes the new proxy for privilege, only with different packaging.
There is another issue to face. If every student has a tutor at all hours, what does authentic assessment look like? How do we measure understanding when the line between getting help and getting answers is blurred? This is not a reason to resist change. It is a reason to rethink what we are measuring and why.
What I Got Wrong, and What I Got Right
The shift is cultural as much as it is technological. For years, tutoring companies helped reinforce the idea that school alone was not enough. Now, AI is challenging that notion and putting powerful learning tools directly in the hands of students. I was right about that.
But the real revolution may not be the end of tutoring. It may be its transformation.
This changes the teacher’s role as well. When information delivery and step by step support are available on demand, teachers become something more valuable. They become learning architects who design rich tasks. They become coaches who know when to push and when to support. They become mentors who help students navigate not only content, but the process of learning itself. The human element does not disappear. It becomes more essential, only with a different focus.
We may soon look back on the tutoring era the way we look at encyclopedias and phone books. Useful for their time, but unnecessary once the world changed. Or we may find that the future looks more like Ector County: AI and humans working together, each amplifying what the other does best.
Maybe what we should have wanted all along was not a system where extra help was a luxury, but one where every student has access to the support they need, when they need it, in the form that works best for them. Whether that form is human, AI, or some combination we have not yet imagined.
The question is not whether this change is coming. The question is whether we will shape it with intention, or let it happen to us.
Thanks to Liz Hill and Andrew Holland with whom I had recent conversations that helped inspire this post.
The image at the top of this post was generated through AI. Various AI tools were used as feedback helpers (for our students this post would be a Yellow assignment – see link to explanation chart) as I edited and refined my thinking
