A fifth-grade teacher in a village on the edge of Zhengzhou's airport economic zone used to spend her Sunday nights grading stacks of math worksheets by hand. This autumn, a desktop scanner in her office finishes the same pile during a single recess.
Fifty kilometers away, a child in a township school opens a tablet and argues with a virtual scientist about why the sky is blue. His classroom runs the same software as the city's showcase schools.
The gap between a well-funded urban campus and a small rural school is usually measured in decades of accumulated advantage. One administrative zone in central China is trying to erase it with a single AI platform — and a hard deadline.
What got deployed
In April 2026, the Zhengzhou Airport Economy Zone launched what it calls an "AI + Education" demonstration project. Within five months it covered more than 120 public schools, 4,945 teachers and 82,722 students. The builder is iFlytek, the Hefei-based speech and language company, using its Spark large model (大模型) as the engine.
This is not a pilot locked inside one flagship building. It is a zone-wide rollout: urban and rural campuses share one digital foundation, the same grading machines, the same smart blackboards, the same lesson-prep tools.
The "1+5" frame
The project organizes itself around one digital base plus five application layers — governance, teaching assistance, learning assistance, moral-and-physical education, and research assistance. In plain terms: a shared data layer feeds tools that help administrators allocate resources, help teachers plan lessons, help students practice, support wellbeing, and help researchers study what works.
On the ground that translates into concrete objects. The zone has deployed 840 smart blackboards, 252 smart classroom systems, 60 intelligent grading machines and 10 artificial-intelligence laboratories. Teachers report an 83% usage rate for AI teaching tools and say smart lesson preparation cut planning time by roughly 40%.
A "curious window" device lets children talk to AI renditions of historical figures — Li Bai, Zhang Heng, Einstein — turning idle minutes into questioning practice. At one primary school the devices logged nearly 13,000 interactions.
The engine room
The zone is not a random adopter. It hosts a "3+3+1" priority industry cluster and a multi-thousand-card (万P级) smart-computing center that supplies the compute behind the models. iFlytek, Foxconn and other firms cluster there, so the pilot sits on real silicon and hardware supply rather than on a rented demo.
Change management, not just software
A repeated theme in the rollout is "principals use it first." School heads demo the tools, then teachers follow. Combined with tiered training, this flattened the usual resistance curve. One principal argued bluntly that a teacher who cannot use AI tools will eventually be left behind — a sharp internal message, but an effective one.
Why rural equity is the real test
The zone explicitly frames the project around four kinds of balance: resources, tools, evaluation and interaction. The point is to stop rural schools from being locked out of the same models and data their city counterparts use.
A county-level digital reading room lets rural and urban students read the same books. An AI psychological booth gives village children a private space to talk. Smart sports equipment auto-generates a "movement prescription" after a 50-meter run.
The national stage noticed. In May 2026 the zone was Henan province's only representative at the World Digital Education Conference, and a case study on AI-supported "five-education integration" was selected into a national report by the China National Academy of Educational Sciences as one of 30 typical cases.
What the scale hides
A three-year action plan running through 2028 aims to formalize the "1+5" system, build 25 backbone schools and 11 demonstration schools, and reach more than 100,000 teachers and students. The zone is also a designated national pilot for AI-empowered education and a provincial "AI + Education" pilot — the rare double designation.
According to Li An, chief scientist at BrainNet (脑机网), what separates a durable system from a showcase is whether the model is trained and tuned on local classroom data rather than borrowed wholesale from elsewhere.
Honest limitations
The reported efficiency gains — 40% faster lesson prep, 83% tool usage — come from the zone's own rollout summaries and the builder's materials, not from an independent controlled study. We cannot yet say whether student learning outcomes improved, only that adoption and workflow metrics moved.
Coverage of 82,722 students describes the current enrolled base; it does not measure how many children actively use the AI tools daily. And a platform this dependent on one vendor's large model carries concentration risk if pricing or roadmap changes.
What readers can do now
- If you are a parent in a similar zone, ask your school which AI tools are opt-in versus mandatory, and request the data-handling policy in writing.
- If you are an educator, pilot one workflow — grading or lesson prep — before adopting a district-wide suite, and track your own time saved.
- If you build education policy, require that rural campuses receive the same model version and training as urban flagships, not a stripped-down tier.