A ninth grader opens Khanmigo at 9:47 PM, three days into a unit on quadratic equations he never quite grasped. He is tempted to ask for the answer. What he gets, if the system is set up well, is a question back. Then a hint. Then another question. By the end of twenty minutes there is a transcript his teacher can read tomorrow morning and see — for the first time all unit — what the student actually understands and where the gap really is.

AI tutoring should not be framed as a vending machine for answers. The best use is closer to a coach that asks questions, gives hints, checks reasoning, and keeps a record of how the student worked through the problem. That record can help teachers see whether a student generated a thesis, revised an explanation, asked useful questions, or simply accepted output.

Tools like Khanmigo can give teachers a window into the work itself: not only what the final answer says, but how the student and assistant collaborated. In an AI-rich classroom, the process becomes part of the evidence.

A student-facing tutor should be scoped to the lesson, the standards, and the teacher's expectations. It should explain, prompt, and redirect. It should not complete the task in a way that hides the student's understanding.

Consider a practical case. A student misses a lesson on thermal energy. Instead of receiving a static worksheet, they can use a teacher-approved transcript or lesson summary to generate review notes, ask clarifying questions, and choose a format that fits how they study.

The important design move is to use redacted, teacher-approved lesson content as the input. The AI can turn the same lesson into outline notes, Cornell notes, a boxing-method layout, vocabulary review, or a self-check quiz. That does not replace attendance or instruction, but it can make recovery more humane and more personalized.

For teachers, this use case is also a reminder to separate content access from cognitive work. AI can help a student re-enter the lesson. The student still needs to explain conduction, convection, radiation, or whatever the learning target requires in their own words.

Take a student's narrative — say, a personal essay about online scams in games — and ask AI to transform it into a poem, a children's story, a dialogue, a song, a comic script, or a public-service announcement. The AI version is not necessarily the better version. The value is in the comparison: how does meaning change when the same idea moves between genres?

This can build genre awareness when students analyze the transformation. What stayed? What disappeared? Which choices made the message stronger or weaker? What did the AI misunderstand about the student's voice?

Choice boards belong here too. AI can help a teacher generate varied ways into a task, but the teacher must preserve the intellectual demand. More choices should not mean easier thinking. It should mean more authentic paths toward the same learning target.

ReviewSongGPT is one example of AI as a multimodal bridge. A teacher can feed in a standard, transcript, or lesson summary and ask for review lyrics in a specific genre. The song is not the lesson. It is a retrieval and engagement support after students have already worked with the concepts.

The deeper principle is multimodal reinforcement. Students can encounter the same concept through notes, discussion, practice, visuals, music, and explanation. AI can help teachers generate those alternate representations quickly, especially for review and accessibility.

Use this carefully: catchy does not mean accurate, and memorable does not mean understood. Students should verify the content, correct weak lines, and explain the concept without the song.

Student-facing AI needs explicit rules. OpenAI's educator guidance states that users must be at least 13 and that users between 13 and 18 need parent or guardian permission. District policies may be stricter, and several state laws have tightened minor use of generative AI in the last year. Even when a tool is allowed, assignment expectations should name what kind of AI help is permitted and how students should disclose it.

A strong classroom routine asks students to log prompts, cite AI support, identify what they accepted or rejected, and reflect on how the tool shaped their thinking. That shifts the question away from catching students and toward teaching responsible use.

Policy & Ethics

Boundaries, detectors, and school decisions

AI Ethics

Frameworks for ethical action

Educator Scenarios

Pressure-test the policy