10 Chinese EdTech Companies to Watch in 2026: AI Tutors, Smart Classrooms, and Learner Autonomy

Imagine two students working on the same mathematics problem. One takes a picture, receives a solution in seconds, and moves to the next exercise. The other receives a clue, explains an attempted strategy, makes a mistake, and corrects it. Both have used educational technology. Only one has necessarily practised the kind of thinking that can eventually make the technology unnecessary.

That distinction is central to China's education technology story in 2026.

Chinese EdTech is no longer adequately described as a collection of after-school tutoring apps. It includes adaptive learning engines, AI-enabled study devices, classroom operating systems, coding environments, robotics kits, teacher tools, and live instruction supported by software. Some of these companies are comparatively young. Others are established businesses reinventing themselves after years of regulatory and market change.

This article examines 10 Chinese EdTech companies to watch in 2026 and asks what their products change for learners, families, teachers, and schools. It is an editorial watchlist, not a verified ranking of the ten fastest-growing companies. Inclusion reflects a distinctive educational use case, a documented product or business development, or an instructive example of market transformation. Sales, user registrations, and funding announcements are not treated as proof of better learning.

Why is Chinese EdTech changing so quickly in 2026?

On April 2, 2026, China's Ministry of Education outlined its latest digital education priorities, emphasizing the integration of artificial intelligence across teaching, learning, and educational management. A broader AI + Education Action Plan, jointly issued by five government departments, envisaged AI literacy across school stages, support for teachers, and technology-enabled personalized learning.

That national direction matters. When schools, families, and companies receive a consistent signal that AI literacy, digital resources, and human–machine collaboration are priorities, demand can shift from isolated tutoring applications toward integrated educational ecosystems.

Yet public policy also highlights the risks. A digital education ethics framework released in May 2026 emphasized human decision-making, appropriate human–AI collaboration, and avoiding overreliance on machines. In other words, China's own policy conversation recognizes that better technology is not automatically better education.

The market's earlier regulatory restructuring, particularly around academic tutoring, is another important backdrop. Today's companies cannot simply assume that the old model of scaling after-school lessons will return. They must find defensible roles in school services, enrichment, learning hardware, personalized software, adult learning, or international markets, while navigating applicable rules.

At a glance: ten Chinese EdTech models worth examining

Company	Core approach	Main audience	The learning question
Squirrel AI	AI-adaptive tutoring and diagnosis	School-age learners, families, centers	Can students explain the strategy without prompts?
17EdTech	School services and AI learning applications	Districts, schools, families	Does personalization improve understanding?
ClassIn	Digital and hybrid learning environments	Teachers, schools, students	Is participation meaningful, not merely measurable?
Codemao	Coding and computing education	Children, teachers, schools	Can learners build and debug independently?
Makeblock	Robotics and project-based STEM	Schools, makers, families	Can students transfer ideas beyond a kit?
Spark Education	Interactive mathematics and reasoning classes	Children and families	Does engagement develop genuine reasoning?
Yuanfudao	AI-enabled learning products and services	Students, parents, teachers	Who controls the learning path?
Zuoyebang	Homework assistance and learning hardware	School-age learners and families	When does a hint become an answer shortcut?
TAL / Xueersi	Learning content, enrichment, smart solutions	Children, families, schools	Do business gains translate into learning gains?
NetEase Youdao	Learning services, AI and study devices	Students and lifelong learners	Which product types actually help learners?

The order is thematic, not a league table. Companies differ substantially in age, funding structure, customer type, and educational scope; comparing their revenue figures as though they were equivalent would be misleading.

1. Squirrel AI: can adaptive tutoring do more than identify a knowledge gap?

A learner who repeatedly fails fraction questions may need a different explanation, but the real issue could be more specific: misunderstood equivalence, weak number sense, or an inability to apply a rule in unfamiliar situations. Human tutors try to locate that hidden gap through questions and observation. Squirrel AI applies AI-driven diagnosis and adaptation to the same broad challenge.

Its proposition is that a system can identify fine-grained areas of knowledge, assign targeted practice, and adapt the next learning step to the learner's performance. The company's technology descriptions emphasize personalized pathways rather than a fixed sequence followed by every student.

The commercial model blends software and tutoring-oriented delivery, including learning centers and related products. Its wider reach and impact claims should be understood as company-reported unless independently validated. A platform can complete millions of diagnostic interactions without demonstrating that children can solve new problems unaided.

The pedagogical opportunity is sharper than the marketing vocabulary suggests. Imagine a student receiving three carefully selected problems after a diagnostic assessment rather than thirty repetitive ones. If the student can explain why the new problems differ and how they solved them, adaptation has reduced unnecessary effort while preserving thought.

The unresolved question: does the system merely optimize the next exercise, or can the learner progressively take ownership of diagnosing their own errors? The strongest outcome would be a student who begins to notice patterns without needing an algorithm to point them out.

2. 17EdTech: the reinvention of school services through AI

Once widely associated with its 17zuoye education platform, 17 Education & Technology Group now illustrates a different branch of the Chinese EdTech market: services for schools alongside consumer-facing AI learning applications.

Its clearest recent business milestone was reported in a September 9, 2026 filing. Second-quarter 2026 revenue reached RMB90.1 million, up from RMB25.4 million in the equivalent 2025 quarter. The company reported net income of RMB1.1 million after a loss a year earlier. Management attributed much of the revenue increase to its Yiqi Aixue consumer AI service, alongside school- and district-level projects.

These are meaningful signs of business momentum, but they are not independently measured learning outcomes. The company's unaudited quarterly figures also should not be mistaken for a complete picture of long-term financial durability.

For a school administrator, the attraction is coordination: teachers can identify needs, families may receive personalized learning support, and education authorities can examine broader patterns. The danger is assuming that more dashboards automatically create better instruction.

A productive question for any pilot is whether a teacher can recognize a misconception, intervene, and later observe a learner overcoming a similar problem without extra automated support. If the answer is yes, analytics may have aided a real teaching decision. If not, sophisticated data may simply describe the same difficulty more precisely.

3. ClassIn: will digital classrooms become better learning communities?

At first glance, ClassIn is not as flashy as an autonomous AI tutor. It focuses on the infrastructure and interaction of learning: online and blended classes, teacher orchestration, digital learning tools, and a common space for educational activity.

That distinction is important. A powerful AI chatbot serves one learner at a time. A well-designed classroom also serves relationships among learners, including explanation, disagreement, peer feedback, and collaborative problem solving.

ClassIn, developed by EEO, provides a platform on which teachers and institutions can organize these experiences. Its growth claims and international reach are presented through company materials and are not evidence, by themselves, of improved achievement. Its competitive landscape also includes conventional learning management systems and video-conferencing platforms, so its educational value must be identified beyond having more features.

Consider a hybrid classroom in which six students propose different solutions to a science task. Technology can help show their working, record their contributions, and support discussion. But the teacher still decides which disagreement reveals a misconception worth exploring.

The crucial measure is not the count of clicks or speaking turns. It is whether learners revise an idea in response to reasons, develop a shared understanding, and later apply that understanding on their own. Digital classroom infrastructure is useful when it makes those processes easier to cultivate.

4. Codemao: teaching children to create technology, not just consume it

Many EdTech products position children as recipients of content. Codemao, built by Shenzhen-based Dianmao Technology, addresses a different question: can students become creators of software and digital projects?

Its offerings encompass programming environments and learning pathways that introduce computational thinking, visual programming, and more advanced coding, with resources for schools and educators. Its education-focused materials frame programming as both a practical skill and a learning domain.

China's interest in school AI literacy gives coding education renewed relevance, but making a colorful game is not automatically the same as understanding an algorithm. A learner can follow a tutorial, paste a solution, and successfully animate a character while learning little about how their code works.

The more valuable experience begins when something fails. Why did the character move twice? Why does the loop never stop? What assumption about the input was incorrect?

The business proposition includes coding instruction, digital platforms, curriculum resources, and institutional partnerships. Large user counts sometimes quoted in promotional materials need clear definitions: registrations, active learners, paying customers, and completed projects are different measures.

The educational test is straightforward. Give the student a novel task after a supported lesson. Can they plan a solution, choose tools, identify a bug, and explain their design decisions? If so, coding technology has helped build transferable capabilities, rather than another form of guided screen consumption.

5. Makeblock: robotics puts learning back into the physical world

Makeblock brings a different dimension to Chinese education technology. Through robotics kits, programming tools, and classroom-oriented STEM equipment, it connects software decisions to physical consequences.

A small robot that fails to follow a path can provide a particularly memorable lesson. Students must confront friction, sensors, timing, measurement, and mechanical limitations. Their program can be logically sensible yet behave unexpectedly once it meets the real world.

That makes robotics promising for project-based learning and computational thinking. Makeblock's mBot2 ecosystem, for example, combines hardware with approaches that can connect visual coding to more advanced programming. Product specifications come from the company and should not be confused with classroom effectiveness studies.

Hardware also creates adoption hurdles. Schools must consider device price, maintenance, spare parts, teacher training, classroom storage, and whether meaningful projects can be completed in the time available.

A robot may produce a spectacular demonstration for visitors; it becomes educational when students formulate hypotheses, record results, correct a design, and defend what changed. One effective assessment is to change the challenge after the demonstration: an unfamiliar obstacle, a different surface, or a modified sensor constraint.

A learner who understands the underlying ideas should be able to adapt. That is the difference between following assembly instructions and acquiring engineering judgment.

6. Spark Education: interactive math with humans still in the loop

Spark Education—associated with the Chinese-origin Huohua Siwei learning business—has built interactive mathematics and reasoning experiences for children, including live instruction and digital activities.

Its Spark Math program illustrates the appeal of a hybrid approach: digital visuals and structured tasks can support teachers who encourage children to talk through problems, spot relationships, and try multiple methods.

The challenge is distinguishing sustained mathematical reasoning from immediate excitement. Children can enthusiastically complete colorful puzzles without fully understanding the underlying concept. Conversely, an enjoyable learning environment can provide the motivation needed for a child to persist with demanding reasoning.

The company addresses families and international markets as well as learners originating in China. Its commercial metrics and reported family reach should be classified as vendor statements unless matched with independent sources. No broad causal conclusion about higher mathematics achievement follows merely from enrollment or class satisfaction.

For parents evaluating the product, two questions are more useful than asking whether a child had fun: What idea can the child explain in their own words? And can they solve a different problem using that idea without the teacher's immediate assistance?

A service that combines engagement with those forms of transfer deserves attention. The educational ambition should not be to maintain a child's dependence on engaging classes forever, but to develop confidence and reasoning that carry into ordinary schoolwork.

7. Yuanfudao: a major tutoring brand adapts to an AI-first market

Yuanfudao is not a young startup. That is precisely why it matters to a 2026 watchlist: established Chinese education brands are repositioning themselves around AI-enabled learning services and wider product ecosystems.

The company describes a combination of digital learning products, educational content, and AI capabilities. For families already familiar with technology-supported study, the promise is convenience: diagnostics, explanations, planning, and individualized resources delivered through a connected service.

But convenience can conflict with productive struggle. If a student receives an almost perfect explanation immediately after encountering difficulty, they may never reveal how they were thinking. This makes the system less able to distinguish a missing concept from a momentary lapse, and the student has less opportunity to practise problem formulation.

There is also a business distinction between subscriptions that provide sustained educational support and short-lived engagement driven by examination urgency. Pricing, accessibility, and subject coverage should be checked for the specific product and region; there is no single stable price that describes an entire company.

The pedagogy worth watching is progressive release of responsibility. A thoughtful tutor might begin by asking the student what they tried, then provide a smaller hint, and eventually withdraw prompts as competence grows. AI personalization is most valuable when students become less dependent on it, not when their entire study routine becomes impossible without it.

8. Zuoyebang: from homework answers to AI learning devices

Zuoyebang is particularly useful for understanding the changing relationship between software and hardware in Chinese EdTech. Known for homework assistance and problem-solving tools, the company has increasingly emphasized dedicated learning devices and AI-guided study experiences.

Chinese reporting on its 2026 product launches described additional smart learning hardware and increasingly integrated tutoring features. Such launches signal a product strategy, but do not establish that the new models are more effective than a teacher, textbook, or conventional learning application.

The original homework-help question remains unresolved: what happens after a child sees the correct solution?

Photo-based problem solving can be beneficial when it helps identify a mistaken step or demonstrate a method that was not accessible in the classroom. It becomes counterproductive when a learner routinely bypasses the effort of recalling, representing, and checking knowledge.

A useful family rule is to ask for an independent attempt before requesting assistance. The next stage might involve comparing the child's attempt with the AI explanation, identifying the precise difference, and solving a new problem without the device.

Device sales can grow while academic independence remains unchanged; they are separate indicators. Schools and parents should also consider content quality, account security, recurring fees, repairs, and age-appropriate usage. The educational advantage lies in better feedback routines, not merely a faster path to a correct answer.

9. TAL Education Group and Xueersi: growth after reinvention

TAL Education Group, associated with the Xueersi brand, is an established public company rather than an emerging startup. Its inclusion shows how important reinvention has become within China's education sector.

According to its April 23, 2026 earnings announcement, TAL recorded US$3.009 billion in revenue for the fiscal year ended February 28, 2026, compared with US$2.250 billion in the previous year, a reported increase of 33.7%. It also reported positive operating income for the year. These figures show commercial scale and recovery; they do not demonstrate that the average student learned 33.7% more.

TAL's portfolio of learning solutions and enrichment offerings represents a broader industry movement away from a single kind of academic tutoring business. It can combine content, instruction, technology and, where offered, smart learning products.

The relevant question for an educator is whether an integrated ecosystem supports a coherent progression of understanding. More content is not always better. A learner who struggles with a foundational idea may benefit more from careful diagnosis and three well-chosen tasks than from a large personalized feed of exercises.

There is a further issue for households: children may already attend school, receive homework, and use several learning applications. Another ecosystem could reduce confusion if thoughtfully integrated—or increase the workload if it simply adds new assignments.

What would count as evidence? Longitudinal performance on unfamiliar tasks, reliable comparisons against suitable alternatives, learner confidence and self-regulation, and transparent reporting of costs and time commitments.

10. NetEase Youdao: AI learning services outperform devices in one recent snapshot

NetEase Youdao offers one of the clearest reminders that the EdTech market is not moving uniformly upward.

In results released on August 20, 2026, Youdao reported second-quarter learning-services revenue of RMB795.6 million, up 20.9% from the same period in 2025. At the same time, smart-device revenue fell 31.5% to RMB86.8 million. Total company revenue increased 3.5% to approximately RMB1.5 billion.

The mixed picture matters more than a headline celebrating AI. Demand for tutoring services, digital tools, and hardware can follow different trajectories. An education company may expand its AI capabilities while a particular hardware category weakens.

Youdao is associated with learning services, dictionary and language tools, AI-based applications, and smart devices. Those functions address different problems. Language learning needs accurate input, repeated retrieval, productive conversation, and feedback; general study tools need to strengthen comprehension, not simply condense difficult material into pleasing summaries.

For an international reader, its example suggests that hybrid business models require careful evaluation product by product. Investors might focus on margins and revenue mix. Educators must also ask whether the tool prompts students to retrieve knowledge, explain an error, or transfer a skill to a new context.

A company may be financially healthy while a particular product is pedagogically mediocre. The reverse is also possible. Conflating these two measures is one of the most common errors in discussions of educational technology.

What really distinguishes these companies?

The ten cases reveal several distinct models, not one Chinese EdTech formula.

AI-adaptive tutoring offers fine-grained practice and rapid feedback, but risks creating answer dependence. Digital classroom systems can support teacher-led collaboration, but cannot guarantee that interaction is intellectually meaningful. Coding and robotics invite students to build and test projects, but lose much of their value when learners only follow instructions. Integrated learning businesses may offer continuity and easier access, but their size and profitability tell us little about outcomes for individual students.

A parent's purchasing question is different from a principal's procurement question. A family may care about affordability, attention, ease of use, curriculum relevance, and screen time. A school leader also needs to examine data governance, accessibility, teacher preparation, integration with existing systems, and evidence across different student groups.

The most important comparison is therefore not between the loudest product launches. It is between the learning processes each design encourages.

How should schools evaluate Chinese EdTech before adopting it?

Suppose an international school is considering a six-week trial of an adaptive mathematics application. The marketing material promises a tailored journey for every student. The school could simply record how many exercises were completed and ask whether the interface was engaging. That would be an incomplete evaluation.

A better pilot would start with a clear question: Can this tool help students identify and correct a specific misconception, then solve an unfamiliar problem without it?

Before implementation, teachers would select an assessment that measures the target skill and an appropriate comparison approach. During the pilot, they could observe whether students try first, request hints productively, and explain why they changed their approach. Afterward, they would use a different problem format to test whether learning transfers beyond the app.

Schools should ask vendors for a concise explanation of how content accuracy is reviewed, what student data is collected, who can access it, where it is stored, how long it is retained, and what choices families have. A responsible school should also confirm age suitability and compliance with the laws and policies governing its own location.

Four educational indicators deserve special attention:

• Independent performance: Can the student complete a new task without automated guidance?

• Metacognitive awareness: Can the student describe their strategy and recognize an error?

• Human agency: Can teachers and families override recommendations when they have good reasons?

• Equity and access: Does the design work across different needs, connectivity levels, languages, and budgets?

The cheapest platform is not necessarily the best value, and the most sophisticated one is not necessarily the safest choice. What matters is whether its benefits can be demonstrated in the context where it will actually be used.

The learning-coaching perspective: teach the learner to outgrow the tool

At Learning Coach, the central educational question is not whether a student has access to an intelligent assistant. It is whether the student becomes more capable of learning independently.

This is consistent with the principles of a coaching posture: work alongside a learner rather than replacing their decisions, set meaningful goals, develop awareness of the learner's own strategies, mobilize personal and environmental resources, and make growing autonomy the objective.

A teacher acting as a coach might turn an AI-generated hint into a short conversation: “What did you try before asking? Which part was uncertain? Can you explain the next step in your own words? What will you do if this problem looks different tomorrow?”

That approach does not reject technology. It gives technology a more demanding role.

Educators and school leaders who want to develop this posture can explore the Learning Coach training pathways and specialties. The important distinction is between discovering the program and assuming that every listed specialization has an immediately available, completed certification pathway; candidates should check current details.

For further reflection on attention and the conditions that shape engagement, the English-language Learning Coach books collection offers relevant educational reading. Books and training may deepen professional understanding, but neither is evidence that a particular EdTech vendor produces better outcomes.

Frequently asked questions

What are the leading EdTech trends in China in 2026?

The clearest trends illustrated by these companies are AI-adaptive learning, integrated study devices, school-facing platforms, programming and robotics education, and learning services that combine software with human instruction. China's 2026 AI + Education policy adds momentum, but adoption is not uniform across products or settings.

Are these the ten fastest-growing Chinese EdTech startups?

No. This is a qualitative watchlist. It intentionally includes younger or evolving businesses and major incumbents whose transformations influence the market. Reliable comparisons would require consistent and current definitions of growth, financial reporting, customer numbers, and product scope—information not equally available for all ten companies.

Is AI tutoring more effective than human tutoring?

There is no sound universal answer. The effectiveness of either approach depends on the learner, subject, design, teacher involvement, assessment, and outcome being measured. Product demonstrations and company-funded announcements cannot replace independent evaluation in a relevant educational setting.

What is the biggest risk of AI homework tools?

One serious risk is answer dependence: students may arrive at correct responses without practising recall, diagnosis, reasoning, or transfer. A useful mitigation is to request an independent attempt first and use AI feedback as a clue to examine rather than a shortcut to copy.

Can schools outside China use these products?

Availability varies by vendor, country, age group, language, licensing arrangements, and applicable data-protection law. International schools should confirm availability directly, evaluate translated or localized content, inspect privacy terms, and run a small pedagogical pilot rather than assume a domestic Chinese product will transfer seamlessly.

The best educational AI should make itself less necessary

China's EdTech sector in 2026 tells a more complex story than the race to build the smartest tutor. It includes teachers coordinating digital classrooms, children debugging robots, families choosing learning devices, and major companies restructuring their offerings around new technologies and regulations.

Some companies can point to meaningful financial momentum. Others offer distinctive pedagogical possibilities without comparable public growth metrics. Across them all, the evidence for independent learning outcomes is still more limited and fragmented than confident marketing headlines may imply.

The test that deserves to shape the next stage of education technology is deceptively simple: after months of using the system, can learners do more by themselves?

If the answer is yes—because students ask better questions, notice their own mistakes, try alternative strategies, collaborate intelligently, and apply knowledge in unfamiliar settings—technology has served education.

If the answer is only that they can complete more tasks while the system remains beside them, the industry still has important work to do.

Source note: This editorial review was checked on October 16, 2026. Company product descriptions and financial releases are attributed to their providers; independent classroom impact has not been established for every product featured. Key public references include China's 2026 digital education priorities, the AI + Education Action Plan, the digital education ethics framework, TAL's April 2026 earnings, 17EdTech's September 2026 filing, and Youdao's August 2026 results.