The pitch decks all sound the same. Every vendor promises personalised, one-to-one tutoring at scale, and every district buyer has to figure out which claims survive a pilot. The market is real: the global AI tutors market is projected to reach about USD 2.75 billion in 2026 and grow at roughly 30 percent a year, with K-12 holding the largest single share.
This guide is written for the people who actually sign the contract: school IT admins, district technology buyers, and EdTech product managers. You will get a fair, sourced comparison of the leading AI tutoring software for school use, a free-versus-paid breakdown, a six-check evaluation framework, and an honest read on when a custom build beats anything off the shelf. Published 3 June 2026. Last updated 3 June 2026.
- "AI tutoring software" covers four different products: student-facing tutors, teacher assistants, adaptive courseware, and reading tutors. Match the category to your need before comparing brands.
- Khanmigo is free for US teachers and $15 per student per year for districts; MagicSchool district plans start around $3,000 per year. Free tiers exist but cap users or features.
- FERPA is the dealbreaker. The 2025 rules shifted student-data consent from opt-out to opt-in, so confirm compliance in writing before piloting.
- Build custom when off-the-shelf tools cannot align to your curriculum, integrate with your LMS, support your languages, or meet your data-residency rules.
What counts as AI tutoring software for school (and what doesn't)
The first mistake districts make is treating one category of tool as if it were another. A lesson-planning assistant is not a student tutor. An adaptive math platform is not a reading coach. Sorting the market into four buckets saves you from comparing products that were never meant to do the same job.
Most districts need more than one category, which is the first hint that a single purchase rarely covers a whole system. Keep that taxonomy in mind as the named comparison gets specific.
The four AI tutoring platforms K-12 districts shortlist in 2026
These four come up in almost every district evaluation we see. Each is strong at one job and weaker at others. The honest summary: pick for your primary use case, because none of them does all four jobs well.
Khanmigo (Khan Academy) is the most established student-facing tutor. It guides learners through problems using a Socratic method tied to Khan Academy's content, and it reports more than 700,000 K-12 students and 380-plus district partners in the 2024 to 2025 school year.
MagicSchool is built for teachers, with 80-plus tools for lesson planning, differentiation, and IEP drafting, and reports adoption by more than 2 million teachers. Carnegie Learning's MATHia is adaptive math courseware with independent studies citing a 12 to 15 percent improvement in math proficiency versus traditional instruction.
Amira Learning is a reading tutor for early grades, fully compliant with FERPA, COPPA, and SOC 2.
Two more names surface in shortlists worth a quick note. Google's Socratic is a free homework-help app that explains concepts, useful for individual students but not built for district administration or rostering. Squirrel AI, widely deployed in Asia, is an adaptive learning system focused on granular skill mapping rather than open-ended tutoring. Neither replaces the four above for a US district rollout, but both can fill a specific gap if your need is narrow.
For a fuller picture of where autonomous education tools are heading, Third Rock Techkno's analysis of AI agents in education is a useful companion read. The pattern it documents holds here: best-of-breed tools win narrow jobs, but they fragment when a district needs one coherent system. A reading tutor, a math platform, a teacher assistant, and a general tutor mean four logins, four data agreements, and four invoices, which is exactly the cost that pushes larger districts toward consolidation.
Third Rock Techkno helps schools run structured pilots and avoid the fragmentation trap. Talk to our EdTech team →
AI tutor app for students: free vs paid in 2026
The free-versus-paid question is where most procurement conversations stall. Free tiers are real, but they almost always cap the number of users, lock the admin controls, or strip the integrations a district actually needs. Here is the honest breakdown.
Free works for a single classroom or a short trial. Khanmigo is genuinely free for US teachers, and MagicSchool's free plan covers a small group of users. The moment you need single sign-on, roster syncing, district dashboards, or a signed data agreement, you are on a paid plan. Treat any free tier as a pilot tool, not a deployment plan, and budget for the paid step before your teachers fall in love with the trial.
"The optimal model is human-AI hybrid vigor, where teachers continue to play an essential role in guiding students' use of the tool."— Brookings Institution, research review on generative AI in tutoring, 2024
How to evaluate AI tutoring software for your district: a six-check framework
Price and brand recognition are the last things to check, not the first. Run every shortlisted product through these six checks in order. A tool that fails check one or two should not reach a pilot, regardless of how good the demo looked.
One check deserves extra weight. According to the Student Data Privacy Consortium, 73 percent of educational institutions report difficulty maintaining FERPA compliance while adopting new technology. If a vendor cannot answer check one cleanly, the other five do not matter. For the wider context on how schools are modernising responsibly, see Third Rock Techkno's guide to digital transformation in education.
We help districts turn this framework into a weighted scorecard and run side-by-side pilots. Get help shortlisting →
When off-the-shelf doesn't fit: building a custom AI tutoring platform
Most districts should buy. Off-the-shelf tools are cheaper, faster to deploy, and good enough for common subjects. The build conversation starts only when a real constraint makes buying impossible, and in our work building custom AI tutoring platforms, four constraints come up again and again.
The first is curriculum that no vendor matches, such as a regional or faith-based syllabus. The second is data residency rules that forbid sending student data to a US cloud. The third is languages and dialects the commercial tools do not support well. The fourth is the need to fold tutoring into a district's own app rather than send students to yet another login. When two or more of these are true, a custom build usually wins on total cost over a three-year horizon.
A custom build is less daunting than it sounds. The common pattern is a retrieval layer over your own curriculum (so the tutor answers from your approved content, not the open web), safety guardrails tuned for your age groups, and direct integration with your existing LMS and student information system.
That keeps student data inside your chosen cloud and lets the tutor speak in the languages your families use. We have found districts underestimate how much teacher trust this earns: staff adopt a tutor faster when it teaches from materials they recognise.
This is the work Third Rock Techkno does through its education software development practice and its Learnly AI tooling, which turns curriculum content into tutoring, lessons, and assessments. The point is not that custom always wins. It is that when a commercial AI tutor app for students cannot meet your curriculum, language, or privacy rules, owning the platform is cheaper than fighting the tool every term.
What to shortlist this budget cycle
Start with your primary use case, not the brand everyone mentions. If you want low-cost cross-subject tutoring, pilot Khanmigo. If teacher workload is the pain, pilot MagicSchool. If math scores are the target, look hard at MATHia, and if early reading is the gap, trial Amira.
Run every option through the six checks, weight FERPA highest, and only open the build conversation if a real constraint rules buying out. Score two or three finalists side by side with the same rubric, and insist on a paid pilot with real student cohorts before any multi-year commitment. The best AI tutoring software for school is the one that fits your curriculum, your data rules, and your budget at full scale, not the one with the loudest demo.
