woman in black long sleeve shirt sitting in front of silver macbook

Should AI Replace Your Child’s Tutor?

With AI tutors set to reach UK schools by 2027, experts at Tutorperch ask the question is simple: can AI replace tutors?

Sal Khan thought AI tutors would close the achievement gap, but two years into Khanmigo, he tells Chalkbeat: “For a lot of students, it was a non event.” This article explores what AI tutoring can do — and what it still cannot.

Students who practised maths with unguarded ChatGPT performed 17% worse on a no AI exam than students who practised without it. Bastani et al., PNAS, 2025

What ChatGPT can’t see when a child says “yes, I get it”

A Year 10 student works through a quadratic with ChatGPT. After a few exchanges, the model says, “Great work, you’ve got it!” The student feels reassured — and fails Friday’s test.

The model cannot tell whether the child truly understands. It’s trained to be agreeable, and the child has learned that saying “I get it” ends the questioning.

This is what researchers call sycophancy, and it is not a bug a future model release will patch. Chuck Arvin’s Check My Work? Measuring Sycophancy in a Simulated Educational Context (USC, KDD Workshop 2025) tested five OpenAI models against 14,042 multiple choice questions from the MMLU benchmark. When the student suggested an incorrect answer in the prompt, accuracy degraded by up to 15 percentage points. The bias is stronger in smaller models, with an effect of up to 30% for the GPT 4.1 nano model versus 8% for the GPT 4o model, and the models shifted towards the student’s suggestion regardless of whether the student was right.

The implication is uncomfortable: a child who half understands a topic may leave believing a wrong but plausible answer rather than being corrected.

Human connection underpins every effective tutoring relationship. A tutor who truly knows their student can deploy the pedagogical approaches that produce the deepest learning: Socratic questioning, metacognitive coaching, formative dialogue. Socratic questioning needs a student willing to think aloud and share half formed ideas. Metacognitive development needs the safety to say, “I don’t know where I’m going wrong.” Working in the zone of proximal development only succeeds when a tutor knows exactly where a student’s edge is — and when to push it. That kind of professional precision is relational before it is technical. For parents choosing a tutor, human connection is the hardest quality to assess from a profile and the most important one to get right.

Where AI actually does win

AI does offer learning benefits. AI clearly helps in specific areas, which may explain the government’s plan for AI tools to support up to 450,000 disadvantaged pupils.

For most GCSE level conceptual questions it is competent.

• Vocabulary and spelling drills. Discrete, retrievable items plus spaced practice. An AI assisted flashcard system handles this well.

• Python and code debugging. LLMs read stack traces and explain bugs cleanly for a self taught teenager learning to code.

• Fast essay feedback: three drafts in twenty minutes instead of three weeks. But a trained human teacher gives better feedback. Steiss et al. (Learning and Instruction, 2024) scored 200 pieces of human generated formative feedback against 200 pieces of AI generated feedback on the same essays, and the human feedback was better on 80% of the elements measured. The differences in overall quality were modest set against the time savings, varied with essay score, and the paper concludes that ChatGPT has potential as an evaluative tool given the trade off between quality and time.

In short, AI raises the floor of independent study.

Two RCTs that bracket the question

Two recent randomised trials show what happens when AI is used in developing learning.

The strongest causal evidence in favour of human tutor plus AI is the Stanford Tutor CoPilot RCT (Wang, Ribeiro, Robinson, Loeb and Demszky, October 2024). 900 tutors and 1,800 K 12 students from under served communities; the AI assistant gave tutors real time pedagogical suggestions during live sessions. The result: students of tutors using the assistant were 4 percentage points more likely to master a topic (p<0.01). The subgroup result mattered more than the headline. Students of the lower rated tutors gained 9 percentage points. The weakest tutors with an AI co pilot beat their unaided peers.

This shows what AI is for: not replacing tutors, but helping them decide the next move.

Assessment of learning, and assessment for learning

One subtle limitation of AI marking is its reliance on mark schemes, which are blunter than they appear. A student may deploy correct reasoning through an unconventional route, use domain appropriate vocabulary in an unexpected way, or produce a response technically outside the accepted answer but demonstrating genuine understanding. A trained examiner, and certainly an experienced tutor, can recognise this and credit it accordingly. AI systems optimised for consistency tend to reward proximity to expected phrasing rather than quality of thought. In creative subjects, extended writing, and higher order scientific reasoning, this matters enormously. The mark scheme is a scaffold for human judgement, not a substitute for it, and applying it well has always required interpretive flexibility that remains, for now, distinctly human.

AI marks quickly and consistently, but marking is not understanding. An algorithm can identify a wrong answer; it cannot easily determine whether the error stems from a conceptual gap, a misread question, careless arithmetic, or the particular way a topic was explained three weeks ago. Effective tutors mark formatively, not simply scoring responses but interrogating them, asking follow up questions, and using errors as diagnostic tools to

redirect teaching. This is the difference between assessment of learning and assessment for learning. AI tools are increasingly capable at the former, but the latter requires the kind of contextual, relational knowledge that builds over time between a tutor and student. AI tells a student they’re wrong; a tutor tells them why and what to do next.

From the evidence we can conclude:

Between sessions, AI raises the floor. Vocabulary drilling, retrieval practice, code debugging, first pass essay drafts, the 10pm panic. Free or near free, always available.

At the session, a human tutor raises the ceiling. Diagnostic “do you actually understand this?”, exam board specific marking, motivation, accountability for a student who will not open the app, pastoral judgement during an exam anxiety meltdown, and — in some tutors — an extra eye for safeguarding.

Behind the session, the tutor uses AI as a force multiplier. Generating problem variants at the right difficulty, handling some marking workload, offering alternative teaching ideas, drafting personalised practice, speeding up preparation by creating a quick worksheet or reorganising a sorting task.

Sal Khan’s April 2026 admission is striking because Khanmigo had every advantage: the brand, the OpenAI partnership, his own voice, and a year of teacher feedback. Kristen DiCerbo, Khan Academy’s Chief Learning Officer, told Chalkbeat: “So far I am not seeing the revolution in education.”

The tools are improving, but there is still no evidence that AI can replace the relational, adaptive, diagnostic work of a skilled tutor. When it comes, it will look like a human tutor using AI not an empty chair and a chat window.

Previous post University of Sussex achieves end-to-end spend management with Proactis and Unit4
three children holding hands standing on grasses Next post Welsh Government must deliver greater ALN funding across Wales, says leading Welsh charity