AI tutor vs AI coach: what's the difference for schools?

An AI tutor helps a student learn something specific. It explains, asks checking questions and corrects mistakes until the student understands the material. An AI coach helps a student manage their own thinking on open-ended work. It asks questions, prompts planning and reflection, and leaves the decisions with the student. Both can decline to hand over answers. The real difference is what each one is trying to build: a tutor builds knowledge, and a coach builds the student's judgement and process.

For school leaders, the distinction is practical. It shapes which tool fits which task, what you ask vendors and what staff need to see.

Where does the distinction come from?

The clearest version comes from Wharton's Ethan and Lilach Mollick. In Assigning AI: Seven Approaches for Students, with Prompts (2023), they set out seven roles AI can play for students. Two of them are the AI tutor, for increasing knowledge, and the AI coach, for increasing metacognition: helping students reflect on what happened and plan what comes next.

Each role, they argue, has its own benefits and risks. You can't judge an AI tool well until you know which job it is meant to do.

What is an AI tutor?

An AI tutor gives direct instruction on a topic, adjusted to the student. Good tutoring, the Mollicks stress, is more than explaining: good tutors get students to restate ideas, explain their reasoning and generate their own responses.

Many tools now aim to work this way. Khan Academy describes Khanmigo as an always-available tutor that guides students to find answers themselves. General chatbots have added learning modes too. OpenAI launched study mode in ChatGPT in July 2025; it asks questions to test a student's understanding. Google's Guided Learning in Gemini breaks problems down step by step and uses open-ended questions. Anthropic describes the learning mode in Claude for Education as guiding students' reasoning rather than providing answers.

By the Mollicks' definitions, these are mostly tutors: the destination is a correct understanding of the material. The main risk the Mollicks name is confabulation: a tool that makes up plausible but wrong answers.

What is an AI coach?

An AI coach works on the student's process rather than the content. It might help a student reflect after a team project, or run a pre-mortem before one, imagining how the project could fail and planning around it. The Mollicks' paper includes prompts for both.

A 2026 preprint on AI coaching for engineering students draws the same line. Its authors describe coaching as facilitative guidance that develops metacognition, motivation and the capacity to navigate complexity, while tutoring focuses on correcting problems.

The tasks a coach supports often have no single right answer; a business plan or a comic book can go many good ways. So the coach's job is to keep the student in charge. The Mollicks' guidance tells students the same: they lead the process.

AI tutor vs AI coach: how do they compare?

The two are not rivals. Most students need both knowledge and the judgement to use it. What matters is that a tool is clear about which job it does.

Isn't "won't give the answer" enough?

It's a start, not the whole story.

Consider a field experiment with nearly 1,000 high school maths students in Turkey, published in PNAS in 2025. One group practised with a standard ChatGPT-style assistant. Another used a version designed to give hints rather than answers. Both groups did far better during practice. But once the AI was taken away, the first group's exam grades were about 17% lower than those of students who never had it. The hint-based version largely removed that harm, and its students scored about the same as the group with no AI.

Two lessons follow. Unguarded AI can let students skip the thinking that learning depends on. And in that study, withholding answers prevented harm but did not, by itself, add learning. Designing for the student's own effort is the floor, not the ceiling.

There is also the question of who holds the switch. At launch, TechCrunch reported that students could easily switch back to regular ChatGPT, and that OpenAI was not then offering tools for parents or administrators to lock students into study mode.

Singapore's Ministry of Education (MOE) describes a different pattern for the Learning Assistant in its Student Learning Space. In a September 2026 letter, MOE says the tool requires teacher activation, lets teachers set a maximum number of interactions, and uses prompts that encourage students to think and consider different perspectives. Teachers can also review students' conversation logs. The letter describes MOE's own schools and tools, but the principle travels: AI should support learning "without replacing the thinking and effort through which meaningful learning takes place".

What does a coach look like in practice?

An illustrative example, constructed for this post rather than taken from a real student:

A 15-year-old student writes: "I want my project to be about flooding near our school. What should I do?"

A tutor-style reply explains how urban flooding works, then checks understanding with a question about drainage.

A coach-style reply asks instead: "What have you noticed yourself when it floods? Who is most affected? What would a good result look like by the end of term?" Later it might ask: "What could you test this week, and what might stop it working?"

Both replies are useful. Only the second leaves the student holding the project.

Where does coaching get harder?

Coaching is the more ambitious role. The Mollicks warn that an AI coach may give bad advice, pick up a student's anxious or curt tone, or lose track of a process. They suggest using it in class, with oversight. The engineering preprint goes further: "when AI shifts from tutoring to coaching, the risks multiply". Its student participants accepted AI for technical problem-solving but were sceptical of its capacity for moral, emotional and contextual judgement. The authors argue for keeping a human expert in the loop.

The evidence is also thinner: the coaching sources here are a framework paper and an exploratory preprint, not controlled trials in schools. You should hear that plainly from any vendor, us included.

What should school leaders ask?

These questions separate a tutor from a coach, and a careful design from a casual one:

1. What is it for? Knowledge of a topic, or the student's process on open-ended work?
2. What happens when a student asks for the answer? And the fourth time they ask?
3. Who turns it on? The student, a teacher or the school?
4. What can staff see? And how does a worrying conversation reach an adult?
5. Where does the student's thinking end up? Can they show the decisions, drafts and failed attempts behind their work?
6. How does it handle a coach's known failure modes, such as bad advice or getting stuck?

Where does The Infinity Lab fit?

We are building The Infinity Lab as a coach. It is a digital AI lab where students learn by doing, through a simple loop: Learn, Challenge, Share. Short, interactive courses build understanding. Real-world projects, such as launching a business idea, put that understanding to work. More projects, like publishing a comic book or tackling a local environmental issue, are coming in the pilot. So is the Share stage, where students show what they made, including the attempts that didn't work.

At the heart of each step is what we call the Play: a guided AI conversation designed to coach a student through a task instead of completing it for them. Our coach aims to ask the next good question. It is designed not to hand over a substantive answer on the first pass, so students work through several rounds of their own thinking first. It is also designed to run in an environment your school manages, with guardrails, safeguarding alerts and visibility for staff.

We are pre-launch, so we have no results to report. Founding Partner Schools will help us shape the rollout, and help us study what AI-assisted learning does for students' skills and interests. You can read more about why we started The Infinity Lab and what we offer schools.

If you are working out how AI should fit into learning at your school, we would be glad to compare notes. Register your school's interest.