AI Tutor or Online School? What ₹500 a Month and ₹60,000 a Year Actually Buy in India

An AI tutor buys explanation and practice, on demand, for ₹200 to ₹2,000 a month. An online school buys a timetable, a teacher who notices, a peer group and — the part that actually decides it — a route to a Class 10 certificate, for ₹45,000 to ₹2.2 lakh a year. They are not competing products. One of them cannot register your child for a board exam.

That is the short answer. The longer one matters, because the honest case for AI tutoring in 2026 is stronger than sceptics allow and much narrower than the marketing claims, and because a growing number of Indian families are quietly running both.

The two price points, side by side

AI tutoring in India is cheap because it is a general-purpose subscription being repurposed for schoolwork, not a product built and priced for a child's academic year. The figures below are vendor-published prices, gathered in September 2026 and used here strictly as prices — not as any indication of teaching quality. Subscription pricing in this category changes every few months, so check the provider's own page before you pay.

Against that, a full online school runs from about ₹30,000 a year at the budget end to ₹2.2 lakh at the small-pod end, with mid-tier programmes clustering around ₹45,000 to ₹80,000. We break the ladder down rung by rung in online school fees in India. The ratio is roughly ten to one — which is exactly why the question gets asked, and exactly why the answer is not a matter of value for money.

What AI tutoring genuinely does well now

Four things, and they are real. Anyone dismissing AI tutors wholesale has not watched a struggling fourteen-year-old use one well.

• Unlimited patient drilling. It will set the twentieth quadratic equation without sighing, and the fortieth. No human tutor at ₹500 a month will.
• Instant feedback. The gap between an error and its correction collapses from a week — the usual turnaround on a school notebook — to seconds, which is the single most useful thing about it.
• Explanation pitched at the learner. Ask it to explain electromagnetic induction again, simpler, in Hindi, using a cricket analogy, and it will, without the child feeling stupid for asking four times.
• Availability at 11pm. The night before an exam, at a hostel, in a town with no good maths tutor, it is there. Access has always been India’s binding constraint in tutoring, and this genuinely loosens it.

Notice what those four have in common: they all assume a learner who already knows what to study, is motivated to study it, and is honest with themselves about whether they understood. That assumption is doing enormous work, and it is where the trouble starts.

What the evidence actually shows — and what it does not

The evidence base for AI tutoring is three years old, small, almost entirely outside India, and mostly measured over weeks rather than school years. Within those limits, three studies are worth knowing, and none of them comes from a company selling tutoring.

Read together, those three say something coherent. AI tutoring works when it is designed to withhold answers and provoke thinking, and when an adult is in the loop. It backfires when it is a general chatbot a child can ask for the answer, because practice performance goes up while actual learning goes down — the child feels competent and is not. The Turkish result is the one to hold onto: a tool that makes homework look better while making the exam go worse is worse than no tool.

Be equally clear about what nobody has shown. There is no published evidence that an AI tutor substitutes for schooling over a full academic year. There is almost no Indian evidence at all, and India's boards, languages and exam formats are different enough that importing effect sizes would be guesswork. If a provider tells you AI tutoring has been proven to raise board results in India, ask which study, by whom, published where — and treat a company blog post about its own users as marketing, not evidence.

What an AI tutor cannot do at any price

This list is not about today's model being weak. These are structural gaps, and a better model next year closes none of them.

• Accountability. Nothing happens if your child does not open the app on Tuesday. No attendance register, no teacher asking where she was, no consequence. For a self-driven sixteen-year-old this is freedom; for most thirteen-year-olds it is the whole problem.
• A peer group. No classmates, no group project, no one to compare notes with at 9pm, no friendship. Adolescents learn a great deal from each other, and a chatbot supplies none of it.
• Noticing that a child is quietly struggling or unhappy. A teacher who sees the same child four days a week spots the withdrawal, the avoided subject, the changed handwriting. An AI tutor sees a session that did not start.
• A credential. No AI system registers a candidate with any board, issues a marksheet, or produces a transcript any Indian college recognises.
• Board registration and exam logistics. Forms, deadlines, subject choices, practical exams, exam centres — all human, all yours.
• Anything a parent cannot supervise. Unsupervised use is precisely the condition under which the Turkish trial measured harm.

An online school supplies all six, and that — not teaching quality — is what the extra ₹60,000 is actually buying. Whether it supplies them well varies enormously; small live cohorts do more of the noticing than a batch of thirty, which is why models like StayQrious's eight-learner pods, and the smaller-batch tiers at other providers, cost what they do. But even a mediocre school does the six things a chatbot structurally cannot.

The credential gap is the decisive point

Here is the hard edge of the comparison. In India, the Class 10 and Class 12 certificate is the gateway to almost everything that follows: Class 11 admission, undergraduate admission, CUET, competitive exams, equivalence for study abroad, most government eligibility. Those certificates are issued by boards — NIOS, CBSE, a state board, Cambridge or another international board — and by nobody else. An AI tutor is not a board, cannot become one, and does not partner with one.

So the AI-only route is not a cheaper version of schooling. It is a study aid with no destination attached. Families who use it as a full substitute discover the gap in Class 9 or 10, when the registration deadline is the thing that matters and there is no institution behind them. Our guide to NIOS Class 10 and 12 registration covers the actual paperwork, and the private candidate route with CBSE and ICSE covers the alternative.

The good news is that the credential is cheap if you separate it from the teaching. NIOS is a Government of India open-schooling board and registers learners directly, subject to its age and eligibility rules. Its Guwahati regional centre lists Secondary admission for five subjects at ₹1,800 for a general-category male candidate, ₹1,450 for a female candidate and ₹1,200 for exempted categories, with ₹720 for each additional subject and examination fees charged separately. Published fee schedules differ between NIOS pages and sessions, so confirm the current figures at sdmis.nios.ac.in before you plan around them.

The hybrid pattern Indian families are actually landing on

Read that table again and the sensible combination is obvious, which is why so many families have arrived at it independently: buy the credential from a board, buy structure from whichever route your child's temperament needs, and use AI for the drilling in between. In practice it takes three shapes.

1. NIOS plus AI, for the self-driven teenager. Register directly with NIOS, study from board material, use an AI tutor for practice, doubt-clearing and explanation. Total cost roughly ₹5,000 to ₹30,000 a year. It demands real self-discipline and an involved parent, and it supplies no peer group — our comparison of an online school versus NIOS self-study sets out the trade-off honestly.
2. Online school plus AI, for a child who needs structure. The school supplies the timetable, teacher, cohort and board route; AI absorbs the endless repetition the teacher has no time for. Roughly ₹35,000 to ₹90,000 a year for most mid-tier programmes, plus a subscription.
3. Mainstream school plus AI, replacing coaching. The largest group in practice: the child stays in a regular school and an AI subscription substitutes for some of a ₹3,000-a-month tuition class. Keep the human tutor for the subject the child is actually failing — that is where accountability matters most.

In all three, AI is the cheapest component and never the load-bearing one. If removing the subscription tomorrow would collapse your child's education, the structure underneath is too thin.

How to read an AI tutor’s outcome claims

This category markets aggressively and is almost entirely unregulated. Five questions separate a defensible claim from a fabricated one.

1. Who ran the study? If the company ran it on its own users, it is a testimonial with a percentage sign attached, not evidence.
2. Compared with what? “Students improved 40%” means nothing without a control group who did not use the product.
3. Measured how, and when? Improvement on in-app practice is the exact metric the Turkish trial showed can rise while real learning falls. Ask for an independent test taken without the tool.
4. On which board, in which language, at which grade? An effect measured on American high-schoolers says little about a Class 9 student writing CBSE Science in Hindi.
5. Is a human in the loop? Every positive result so far involved either a teacher running the sessions or a tutor deliberately engineered to refuse answers. A bare chatbot has no such evidence behind it.

And two specific red flags. Treat any guaranteed score or rank improvement as a reason to walk away — no legitimate education provider can promise it. Treat “personalised AI learning path” as a description of a feature, not a result; it tells you the software branches, not that anybody learned more.

An AI tutor is the cheapest good thing you can add to a child’s education, and the most expensive thing to mistake for the education itself.

So which should you choose?

If your child needs a credential — and in India that is nearly every child — the board route is not optional and the only real question is what structure you buy around it. Choose a full online school when your child needs a timetable, when nobody is free to supervise learning through the day, or when you want the exam logistics handled; how online schools compare with home-based routes is worth reading before you commit to any provider's recognition claims. Choose NIOS plus AI when your teenager is genuinely self-driven and an engaged adult is around. Add an AI subscription in either case; it is the best ₹400 a month in Indian education right now, on the strict condition that it is supervised and is not carrying the weight.

One caution before you act on any of this: board rules, registration windows and eligibility criteria change, and they differ by state. This is not legal or admissions advice — verify your current state and board rules with the board directly before making a decision you cannot reverse.

Frequently asked questions

Can an AI tutor replace school entirely in India?

No. An AI tutor cannot register a child with NIOS, CBSE, a state board or an international board, and cannot issue a certificate. Without a board certificate a child cannot enter Class 11, sit CUET or apply to most Indian colleges. AI can replace some tuition; it cannot replace the credential route.

Is there real evidence that AI tutoring improves results?

There is genuine but early evidence. A World Bank trial in Nigeria found gains of 0.23 to 0.31 standard deviations over six weeks, and a Harvard physics trial found more than double the learning gains of an active-learning class. Both involved a teacher or a purpose-built tutor. A PNAS study of Turkish students found that an unguarded chatbot lowered exam scores by 17%. None of this evidence is Indian, and none covers a full school year.

How much does an AI tutor cost in India compared with an online school?

Consumer AI subscriptions in India run roughly ₹200 to ₹2,000 a month, or about ₹2,400 to ₹24,000 a year. Full online schools run about ₹30,000 to ₹2.2 lakh a year, with mid-tier programmes around ₹45,000 to ₹80,000. NIOS board registration on its own costs a few thousand rupees plus exam fees.

What is the cheapest route to a recognised Class 10 certificate?

Direct NIOS registration is the cheapest recognised route: a regional centre listing puts Secondary admission for five subjects between ₹1,200 and ₹1,800 depending on category, plus separate examination fees. It buys certification only — all teaching, discipline and exam paperwork is yours. Confirm current fees and eligibility at sdmis.nios.ac.in.

Is it safe to let a child use an AI tutor unsupervised?

The strongest evidence says no for academic reasons alone. In the PNAS trial, students with unrestricted chatbot access performed better during practice and 17% worse on the exam taken without it, because they used it as a crutch. Supervision, or a tutor deliberately built to withhold answers, is what separates the positive results from the negative ones.

The ₹500 subscription and the ₹60,000 school are not rivals, and treating them as rivals is what leads families astray. One is a practice tool with strong early evidence and structural limits that will not go away. The other is an institution that can put your child's name on a board register. Buy the register first.