How much does an AI tutor cost? Pricing models compared

How much does an AI tutor cost?

AI tutor cost depends less on the product than on how it is priced. The same student using a tutor 30 minutes a week can cost you $3.90 a month on per-minute pricing at $0.03, a flat fee on a seat license whether they log in or not, or several hundred dollars a term with a human tutor. The only reliable way to compare is to take one realistic usage scenario and price it under every model.

This guide does that for a 100-student cohort, with the arithmetic shown so you can swap in your own numbers. All prices are as of September 2026 and come from each vendor’s or organization’s own page, listed at the end.

What are the four AI tutoring pricing models?

Almost every option you will be quoted fits one of four models. The reference points below are published prices, not negotiated quotes, and they are for different products, so use them to understand the shape of each model rather than to pick a vendor.

Per-seat and per-student licenses

A seat license charges a fixed amount per enrolled student, usually per month or per year, whatever each student does. It is easy to budget and easy to approve. A well-known published consumer price is Khanmigo, Khan Academy’s AI tutor: as of September 2026 its pricing page lists $4 a month or $44 a year for learners and parents, free access for teachers, and quoted pricing for districts. Khanmigo teaches on Khan Academy’s own content, so treat that as a price reference, not a like-for-like quote for tutoring on your material.

The catch with seats is that you pay for enrollment, not learning. Every student on the roster costs the same whether they use the tutor daily or never open it. Before signing, ask what usage the vendor sees across similar customers, whether unused seats can be reassigned, and whether the contract has a minimum seat count or an annual term.

Per-minute usage pricing

Usage pricing charges for each learner-minute of live tutoring. Tutezy, for example, charges $0.03 per learner-minute, or $0.05 while an animated teacher avatar is on screen, from prepaid credits, with no seat licenses or minimums. A student who never logs in costs nothing, which makes pilots and seasonal courses (exam prep, onboarding cohorts) cheap to run. The how it works page explains why a lesson compiled in advance keeps the live cost per minute low.

The trade-off is that the bill grows with use and has no natural ceiling. Ask three questions of any per-minute vendor: what exactly counts as a minute (idle time, reading notes afterwards), which features change the rate, and whether you can cap minutes per student or per course. A fixed session length per course, say 20 minutes, is the simplest cap.

To forecast a per-minute bill, multiply three numbers you can estimate: students who will actually use it, sessions per student per week, and minutes per session. A pilot gives you the first number, which is the hardest one to guess.

What does it cost to build your own AI tutor?

Building your own means paying an AI provider per use and paying engineers to turn the model into a tutor. The API part is easy to estimate. OpenAI’s pricing page, as of September 2026, lists gpt-realtime-2.1 at $32 per million audio input tokens and $64 per million audio output tokens, and $10 and $20 for the mini model. OpenAI’s cost guide says user audio is 1 token per 100 ms and assistant audio 1 token per 50 ms, so a minute of student speech is 600 tokens and a minute of tutor speech is 1,200.

Take a minute in which the tutor speaks for 36 seconds and the student for 24 (a simplifying assumption with no silence):

• Tutor audio: 720 tokens × $64 per million = $0.046
• Student audio: 240 tokens × $32 per million = $0.008
• Total: about $0.054 per minute, or about $0.017 on the mini model

That is before the costs that grow during a conversation. OpenAI’s guide notes that the entire conversation is sent to the model for each response, so later turns cost more (cached input is much cheaper, which helps). Your instructions and course material add text tokens on every turn, and you still need speech hosting, storage and monitoring.

Then there is engineering. The US Bureau of Labor Statistics puts the median software developer wage at $135,980 a year (May 2025), about $11,330 a month before benefits and overhead. Three months of one developer is roughly $34,000 in salary alone, which is more than 30 terms of the per-minute cohort below.

And a voice connection is not yet a tutor. You would still need to turn your material into lessons, keep the model from inventing content, grade answers consistently, decide when to give a hint and when to move on, store each student’s progress, build teacher reports, and re-test all of it every time the provider ships a new model. Building makes sense when the tutor itself is your product, not when you need one for a course.

AI tutor vs human tutor: what does each cost?

For private one-to-one tutoring, the US marketplace Wyzant says the average price of a tutor on its platform is between $35 and $60 an hour (in a December 2024 post), with higher rates in some subjects. At $0.03 a minute, an AI tutor costs $1.80 per student-hour, so the human hour costs roughly 19 to 33 times more.

For schools and districts, the National Student Support Accelerator (NSSA) at Stanford publishes program-level estimates. It says cost estimates for high-impact tutoring programs range from $1,000 to over $3,000 per pupil, and that districts should expect roughly $1,200 to $2,500 per student. Its budget guide lists tutor wages of $15–$20 an hour for novice tutors, $20–$30 for experienced tutors and $40–$60 for master tutors, which the guide converts to about $7–$20 per pupil-hour depending on group size and tutor type.

Price per hour is not the whole comparison. A human tutor brings judgment, motivation and a relationship (our post on Bloom’s 2 sigma problem covers what the research on tutoring says). The practical question is which hours need a human. Our Tutezy vs human tutoring comparison lays out where each one wins.

Worked example: a 100-student cohort for 12 weeks

One hundred students each use a tutor for 30 minutes a week for a 12-week term. That is 100 × 30 × 12 = 36,000 learner-minutes, or 600 student-hours.

Two caveats keep this honest. The per-minute AI figures also carry a small one-time cost to prepare lessons (Tutezy’s pricing page puts a typical 40-slide course at under $2.50). And the human rows buy something the AI rows do not, so the fair reading is not “AI is 20 times cheaper” but “every student can get the practice loop, and the human hours can go where they matter most.”

Per seat vs per minute: where is the break-even?

The average can mislead when usage is uneven. Suppose 30 of your 100 students use the tutor 300 minutes a month and the other 70 use it 30 minutes. At $0.03 a minute that is (30 × 300 + 70 × 30) × $0.03 = $333 a month. A $10 seat for all 100 is $1,000. Per-seat pricing wins only when average use per student is above the break-even, so pull a month of real usage from a pilot before you decide. Our AI tutoring software buyer’s checklist covers the pilot.

Hidden costs to budget

• Content preparation and review. Someone has to supply the material and check what the tutor will teach. Multiply the teacher hours per chapter by your own staff cost.
• Onboarding and platform fees. If the tutor only runs inside the vendor’s app or platform, that platform may have its own price, and students need accounts.
• Minimum commitments. Annual terms, minimum seat counts, prepaid tiers and auto-renewal clauses.
• Overages and caps. What happens above a fair-use limit on a seat plan, or when a per-minute cohort uses more than planned.
• Extras. Avatars, custom voices, extra languages and premium support, sometimes priced separately.
• Regeneration. Some products charge again when you rebuild a lesson after changing the source.
• Taxes. Published prices may exclude sales tax; Khanmigo’s page says its prices do.
• Data export and exit. Whether transcripts and reports export as CSV at no charge, and what happens to them when you leave.

If you are running a tutoring business rather than buying for a school, the same arithmetic drives your margins; see how to scale a tutoring business with AI.