Blog

What if every student had a personal tutor?

Throughout the history of education, the most effective form of instruction has also been the most exclusive. Wealthy families have always hired tutors: Roman aristocrats retained Greek pedagogues, Renaissance nobles employed scholar-tutors, and the British aristocracy sent their children to be educated by private masters long before public schooling existed. The tutorial system at Oxford and Cambridge, where students meet individually or in pairs with a tutor each week, remains the gold standard of university education and the primary reason those institutions produce the outcomes they do.

The pattern holds in modern education. Affluent families spend thousands of pounds on private tutoring, prep school students receive personalised coaching, and the students who perform best on standardised tests disproportionately come from families that can afford one-on-one academic support. A private tutor in the UK costs £25-£60 per hour. In the US, SAT tutoring runs $50-$200 per hour, with premium services charging significantly more. These prices ensure that the students who receive the most effective form of instruction are overwhelmingly those whose families can pay for it.

Benjamin Bloom's 1984 research quantified what everyone intuitively knew: one-on-one tutoring produces dramatically better outcomes than classroom instruction. Students with personal tutors outperformed 98% of conventionally taught students. About 90% of tutored students reached the level that only the top 20% of classroom students achieved. The effect was enormous and consistent across the studies Bloom analysed.

Bloom framed this as a problem because delivering personal tutoring at scale was economically impossible. One tutor per student is not a viable model for any education system. The research proved the intervention worked while demonstrating that the world couldn't afford to provide it. For forty years, the two sigma problem stood as an elegant proof of an inaccessible ideal.

The achievement gap is a support gap

The students who benefit most from personal tutoring are the ones least likely to have access to it.

First-generation university students, whose parents didn't attend university and can't help with coursework, arrive without the informal academic support network that many of their peers take for granted. When they don't understand the week 4 reading, they don't have a parent who can explain it over dinner, a family friend who studied the same subject, or the confidence to approach a professor they find intimidating.

Students from lower-income backgrounds are less likely to have experienced private tutoring at school, less likely to arrive at university with strong study skills, and more likely to be working part-time while studying, which limits their availability for office hours and study groups. The students who need the most support have the least access to it, and the gap compounds over time in exactly the pattern that drives dropout.

Students who work while studying face a specific version of this problem. The lecturer's office hours are 2pm to 4pm on Wednesday, but the student works a shift that finishes at 3pm and can't get to campus until 4pm. The study group meets on Saturday mornings, but the student works weekends. The support is theoretically available but practically inaccessible.

The achievement gap between students from different socioeconomic backgrounds is well-documented, persistent, and resistant to most institutional interventions. What the research consistently shows is that the gap is driven less by differences in ability than by differences in support. Students with support outperform students without it, regardless of background, because support is what prevents the compounding gap from forming and what closes it when it does.

Making support universal

An AI tutor grounded in each student's course materials doesn't replace human tutors, academic advisors, or the broader support infrastructure that universities provide. What it does is make a specific, high-value form of support available to every student regardless of their family's income, their work schedule, or their willingness to ask for help in public.

The first-generation student who doesn't understand the week 4 reading can ask the AI tutor to explain it at 11pm using the lecturer's framing and the assigned text. The working student who can't make office hours can get their questions answered between shifts. The student who's too embarrassed to raise their hand in a lecture of 200 people can ask the same question privately and get an answer grounded in their course materials.

The tutor is patient. It never judges. It never gets tired, never has a bad day, never makes the student feel stupid for asking a question. It explains the same concept ten different ways if the first nine didn't land. It's available at 3am on the night before the exam. It costs less per month than a single hour of private tutoring.

This is not a small thing. The difference between having a personal tutor and not having one is, according to Bloom's research, the difference between performing at the 50th percentile and the 98th percentile. For forty years, that difference was available only to students whose families could afford £30-£100 per hour for one-on-one instruction. At around $10 per student per month, the same quality of personalised, source-grounded academic support is accessible to everyone.

What this means for institutions

Universities that deploy AI tutoring at scale are not just improving academic outcomes. They're making a structural intervention in educational equity.

The widening participation agenda, which most universities in the UK and many in the US have committed to, is fundamentally about ensuring that students from underrepresented backgrounds receive the support they need to succeed. Most widening participation spending goes toward outreach, access programmes, bursaries, and targeted support services. These are all valuable, but they're also expensive, human-dependent, and limited in the number of students they can reach.

An AI tutor that's available to every student, that costs a fraction of a single academic advisor's salary to deploy campus-wide, and that directly addresses the academic mechanism that drives differential outcomes is one of the most cost-effective equity interventions available.

For institutions that measure themselves against retention differentials, attainment gaps between student groups, and the success rates of widening participation cohorts, the data from a one-semester pilot would be informative. If the tool narrows the attainment gap between students with and without prior tutoring access, even modestly, the case for campus-wide deployment becomes compelling on equity grounds alone, independent of the financial retention argument.

The mission case

This piece has deliberately made the equity argument rather than the product argument, because the equity argument is the one that matters most and the one that should drive the decision.

The students who would benefit most from personal tutoring at scale are the students who've never had access to personal tutoring at all. The technology to provide it now exists, at a cost that's negligible relative to institutional budgets. The research supporting its effectiveness spans four decades and includes recent randomised controlled trials.

The question for universities is whether they're willing to make the most effective form of academic support available to all their students, or whether they'll continue to reserve it, structurally if not intentionally, for those who can afford it privately.

Frequently asked questions

Can AI tutoring really match the effectiveness of human tutoring? Current research suggests AI tutoring captures several of the mechanisms that make human tutoring effective: immediate feedback, adaptive pacing, personalised explanations, and active engagement. The 2025 Harvard RCT found students learned roughly twice as much per hour with a well-designed AI tutor. It doesn't yet replicate the emotional intelligence of an excellent human tutor, but for students whose alternative is no tutoring at all, it represents a transformative improvement.

Does this risk reducing investment in human support services? The intent is additive rather than substitutional. AI tutoring handles the academic explanation and retrieval tasks that human tutors spend much of their time on, freeing human advisors and tutors to focus on the complex pastoral, motivational, and personal support that AI cannot provide. The overall support envelope should increase, not decrease.

What about students who don't engage with technology? Most university students are already deeply engaged with technology for their studies. The barrier to adoption is typically relevance rather than technology aversion. A tool that helps with the immediate, specific challenge of understanding this week's reading has a different adoption profile from a generic institutional platform.

How does this fit with existing widening participation programmes? It complements them directly. Outreach programmes get students to university. Bursaries help them afford to stay. The AI tutor helps them succeed academically once they're there. The gap in most widening participation strategies is sustained academic support after enrolment, and that's exactly what this provides.