把 AI 工具带进大陆课堂前，需要先确认什么
学校应就可用性、备案、标识与学生数据提出的问题，以及去哪里获得权威答案。
本文尚未翻译成中文，以下正文为英文。
A tool that works beautifully in a demo can be unusable in a mainland classroom for reasons that have nothing to do with quality. Work through these before the lesson, not after.
1. Can students actually reach it?
The first filter, and the one most often skipped by someone evaluating from outside the mainland. Check it on the network the students will be on, on a school device, at the time of day you plan to teach. A tool that needs a workaround to reach is not a tool you can build a lesson around, whatever its features.
Our directory records this as a field on every listing, but treat our value as a starting point and confirm it yourself.
2. Is the service one that has completed the filing or registration it needs?
Public generative AI services offered in China sit under a filing and registration framework, and an application built on an already-filed model's API may go through a different route than a service running its own model. What matters for a school is that the vendor can tell you, in writing, which one applies to them and give you the number.
If a vendor cannot answer that question, that is your answer.
3. Is generated content labelled?
Rules on labelling AI-generated content took effect on 1 September 2025 and cover explicit and implicit labelling in the situations they define. For a school this has a practical edge as well as a legal one: students should be able to see which parts of a handout came from a machine, because that is the habit you want them to carry into their own work.
4. Where does student work go?
Ask directly: is student-submitted text used for training, how long is it retained, and can it be deleted on request. This is a procurement question with a written answer, not something to determine by reading a marketing page.
The safe default while you wait for that answer: no student names, no student work, no assessment data in any general consumer assistant.
5. Who is accountable in the room?
The teacher. Whatever the tool produces, the person who put it in front of thirty students is answerable for it. That is not a legal observation, it is the reason step 4 of any planning workflow is checking the output rather than trusting it.
Where to get an authoritative answer
Not here. This page is a list of questions worth asking. For an actual determination — particularly for anything student-facing, at scale, or involving personal information — a school needs a qualified opinion under Chinese law, and the vendor needs to be on the record.