Arivihan Series A Raises $10M for AI Tutoring in Small-Town India

A ₹51 bet on India's hinterland

Arivihan, an AI-led exam-preparation startup, has closed a $10 million Series A. The money came from Accel and Prosus Ventures, both of which had already backed the company, along with another $200,000 from existing GSF angel investors.23 The company says the round brings its total funding to more than $15 million. Its last raise was a $4.17 million pre-Series A in July 2025.3 Arivihan was founded in 2022 by IIT-Roorkee alumni and went through Accel's Atoms accelerator, which gave it seed funding in March 2024.2

The round is small next to the edtech megadeals of earlier years. It still tells us a good deal about how investors are thinking about AI tutoring, where the product is going, and how much evidence actually supports the claims made around it.

What the money is for

The reports agree on the main uses of the money. Arivihan wants to enter new states, expand its CBSE board coverage, put more into AI research and local-language support, and spend more on performance marketing and on-the-ground distribution.24 It plans to go deeper in Madhya Pradesh, Uttar Pradesh, Rajasthan and Bihar and to launch new crash courses.2 It currently serves students preparing for state board exams, CBSE and the NEET medical entrance test. About 80% of its subscribers come from Tier III cities and rural areas.26

That geography is the core of the pitch. Prosus investor Dhruv Gupta said the company shows AI can deliver individualised learning at mass-market prices. Accel's Anagh Prasad said AI can now make online education as effective as expensive in-person coaching.34 Neither quote comes with outcome data attached. Both describe what the investors expect AI tutoring to do, not what has been measured.

The pricing pivot behind the growth

The most revealing detail comes from co-founder and CEO Ritesh Singh Chandel. Arivihan used to charge about Rs 3,000 a year for board-exam preparation and had roughly 25,000 paying students last year. It then introduced a Rs 51 annual entry plan, and around 4 to 4.5 lakh students have signed up for it this year.1 Chandel said plainly that the goal was volume. He also said 18% of subscriptions to the cheap plan came from referrals.1

Having roughly 16 times more paying students does not mean revenue grew at the same rate, and the company has not shared revenue or how many users move up to its more expensive plans.1 The latest public figures are from FY25: revenue of Rs 3.14 crore, up from Rs 21.29 lakh in FY24, and a net loss of Rs 4.45 crore.7 FY26 accounts have not been filed yet.5

One fair way to read this is that the Rs 51 plan works like a customer-acquisition cost the company has built into its price. It brings in a very large audience at almost no revenue per user, and the business depends on upselling some of them later. That only works if the cost of serving each extra student is close to zero. This is why the AI side of the product matters so much to the company's economics.

Valuation: insiders marking up their own stake

None of the companies involved announced a valuation. Business outlet Entrackr estimated it rose about 3.3 times to around $60 million.5 An analysis of regulatory filings arrived at a similar figure: a post-money valuation of roughly Rs 570 crore, with Accel and Prosus each putting in about Rs 47.48 crore. According to those filings, Accel India owns about 23.54% after the round and Prosus about 18.57%.9

That analysis also pointed out that the valuation is more than 180 times FY25 sales. Because both lead investors were already shareholders, they are in effect raising the price of their own holdings without an outside lead to test it.9 This doesn't make the price wrong. Fast-growing companies often look expensive against revenue that is a year old. But until FY26 numbers come out, the round is evidence that the existing backers believe in the company, not evidence that the wider market agrees on the price.

Is it really teacherless? The coverage disagrees

The biggest difference between reports is how automated Arivihan actually is. Several articles, following the company's own messaging, describe a platform built on interactive video, instant doubt-solving and AI study plans that does not rely on live teachers.610 Business Standard reported the company's claim that removing live teachers makes it ten times more scalable and cost-effective than traditional coaching.3

The Times of India gave a different account. It reported that current classes combine human teachers and AI. A single live class can include up to 5,000 students and one teacher, while AI answers each student's questions using their past mistakes and practice history.1 It also reported that a true one-on-one AI tutor is still on the roadmap and is not yet a product.1

The Times of India version, which is based on direct comments from the CEO, is the more reliable one. Arivihan today is best understood as a very large live class with an AI layer for answering doubts. It is not yet an autonomous AI tutor. The details matter. One teacher reaching thousands of students is already much cheaper than traditional coaching. Replacing that teacher completely is the riskier step, and the company has not taken it yet.

What the research says about AI tutoring

This distinction matters because research on AI tutoring has become sharply split by product design. A Harvard physics trial found that a custom tutor built on the same teaching principles as the classroom lessons produced much larger learning gains, in less time, than in-class active learning.23 Trials in developing countries have also been encouraging. A World Bank study in Nigeria found an improvement of 0.31 standard deviations, and a Ghana trial of the Rori math tutor found an effect size of 0.36.25

Other results are worse. In a trial with about 1,000 high-school students in Turkey, those who had unrestricted access to a GPT-4 assistant during practice scored 17% lower on a later exam taken without it. The assistant had done the thinking that would normally help students learn.28 A University of Maryland trial published on September 30, 2026, the same day as Arivihan's announcement, involved 2,379 undergraduates. It found that access to a course-integrated AI tutor lowered final grades by 0.37 standard deviations and sharply reduced participation in course activities. First-generation students lost more than twice as much as their peers.2426

That last finding should concern any company selling mainly to first-time learners in rural areas. The context is different: American undergraduates are not Indian board-exam candidates. Even so, the warning is that AI tools can push aside the effort that produces learning, and that this tends to hurt students with the least support most.

Some research fits Arivihan's current hybrid model better. In Stanford's Tutor CoPilot trial, giving AI help to human tutors raised topic mastery by four percentage points, and by up to nine points for students whose tutors were rated lower.21 However, the same study found no significant improvement on end-of-year tests.25 Reviewers have also noted that tutoring effects usually get smaller as programmes grow.28

How strong is Arivihan's own evidence?

Arivihan's outcome claims are weak by these standards. The company said about 5,000 students used its NEET preparation in the latest cycle and three of them ranked in the top 100. It also cited internal data showing that 75% of students who completed five mock tests raised their marks by 25% within 30 days. It did not give the size of that group or compare it with students who did not use the platform.1 Students who finish five mock tests are already self-selected for motivation, so the number shows engagement, not that the product works. Since part of the new money is going to "AI research," a randomised evaluation would be a sensible use of some of it.

A thin market for edtech funding

The round comes at a time when there is little money flowing into Indian edtech. Tracxn data show the sector raised $216 million across 37 rounds in 2026 through September, compared with $219 million across 67 rounds in the same period of 2025.14 Total funding is roughly flat while the number of deals has nearly halved, which suggests investors are putting money into fewer, more targeted bets. By our estimate, Arivihan's round alone is close to 5% of the year's total. Commentators describe investors coming back to the category only for low-cost, AI-first models aimed at underserved students rather than the expensive live-teacher setups of the pandemic years.810

The bottom line

Our view is that Arivihan is a reasonable bet on distribution and cost, not yet a proven bet on learning results. It has shown it can reach students in small towns at a very low price, and that is difficult to do. It has not yet shown that the Rs 51 plan can make money, that the valuation is justified, or that its AI improves learning instead of just making students feel helped. The FY26 accounts, data on how many users upgrade, and any independent outcome study will settle these questions. Recent research suggests that outcome data deserves the closest attention.