AI for Teachers: A Practical Guide to Using It Well

How do teachers use AI?

Teachers who get real value from AI use it for one thing: the rough first draft of the work that eats their evenings. A lesson outline. A set of feedback comments. A parent email that has to sound firm but kind. AI writes the version you'd have stared at a blank page to produce — and you spend your time editing instead of starting.

The pattern shows up in the data, not just the anecdotes — and so does the unevenness. In RAND's 2023–24 survey of American teachers, a quarter of ELA, math, and science teachers were already using AI for instructional planning or teaching, with ELA and science teachers adopting at roughly twice the rate of math and elementary teachers. Teachers who stick with it weekly report saving about six hours a week (the numbers, and the prompts behind them, are in our AI lesson planning guide). The loudest conversations about AI in education are still about the wrong thing. The question isn't whether a machine can teach. It can't. The question is how much of a teacher's week is spent on work that isn't teaching — and how much of that you can hand off without lowering your standards.

Where AI actually helps

Six places the drafting handoff works, each with a starter you can copy and adapt (all of these run in any chatbot — for a fuller library organized by job, see ChatGPT for teachers):

• Planning. Turn a standard, a grade, and a time limit into a first-draft lesson in minutes. This one's deep enough that we wrote a separate AI lesson planning guide with prompt examples.
• Differentiation. "Here's the article my 7th-grade science class is reading on plate tectonics [paste]. Rewrite it at a 5th-grade reading level, keep the six key terms with a margin glossary, and give me three text-dependent questions for each version." (When the task requires citing the original text, scaffold around it instead of rewriting it — our guide to differentiated instruction with AI shows why.)
• Feedback. "Here's my 4-point rubric for a lab conclusion [paste]. Write six reusable comments per band: two on the claim, two on use of data, two on reasoning. Sound like a teacher, not a report card." You're editing feedback instead of writing all of it from scratch — and the comments stay yours, because you tune the bank once instead of retyping variations forever. (The full workflow — including where the grade decision stays yours — is in our guide to AI grading and feedback.)
• Materials. "Write 10 practice problems on [skill] in the same format as these [paste two], ordered easiest to hardest, with a worked answer key."
• Communication. "Draft a note to families about [the science fair / a missing-work pattern / a field trip]. Warm but direct, under 150 words, no education jargon." One caution on translations: AI can draft a family letter in Spanish or Vietnamese in seconds, and a mistranslated nuance in a note home is one of the highest-stakes errors on this list. For anything consequential, have a bilingual colleague or your district's interpreter service check it.
• Admin. Meeting notes, permission slips, the endless small writing. For recommendation letters, use AI for the structure, never the student: "Give me a strong structure for a STEM-program recommendation letter — what to lead with, what admissions readers skim, what to cut." Then write the student-specific content yourself; a rec letter is the one document where generic is fatal.

None of these are the job. They're the tax on the job. AI is good at the tax.

What to keep in your own hands

AI produces confident drafts with no sense of when it's wrong, so three things stay yours:

• Accuracy. Read everything. AI states wrong facts with the same fluency as right ones — and in a classroom, a confident error is expensive. The risk scales with precision: discussion questions fail gracefully; answer keys don't.
• Judgment about your students. A tool doesn't know who's had a hard week, who needs a challenge, or what landed yesterday. You do.
• Voice. When AI writes to a family or a student, edit it until it sounds like you. A generic email erodes the trust a specific one builds.
• Grades. AI can draft the comment; it doesn't decide the number. First-pass feedback is a drafting task — the grade of record is a judgment call you sign your name to, and districts writing AI policy are increasingly drawing exactly that line.

What about students using AI?

You can't write honestly about AI for teachers and skip this: the same tool drafting your lesson is writing some of your students' essays, and plenty of teachers feel the hypocrisy question in their gut — can I use it for my work while telling them not to use it for theirs?

Yes, and the distinction is worth saying out loud — to yourself and to your students. You're using AI as a drafting tool with your professional judgment layered on top: you review, verify, revise, and take responsibility for the result. That's not the workaround — it's the exact skill students need to learn, and many of them currently use AI the other way, as a replacement for the thinking rather than a draft for it. "Here's how I used AI to plan today's lesson, and here's what I changed and why" is a better AI-literacy lesson than any policy slide.

The classroom-management side of student AI use — detection, assignment redesign, what to do about the essay that smells like a chatbot — is its own topic and deserves its own guide. The short version: detection tools are unreliable enough that redesigning the task beats policing the output.

What about student privacy?

If one section of this guide deserves a staff-meeting slide, it's this one. General-purpose AI tools are not, by default, safe places for student data. The baseline rules:

• Keep identifying information out of prompts — no names, grades, IEP or 504 details, or anything that could identify a child. Watch the subtle version too: "my one student who uses a wheelchair and struggles with…" identifies a student as surely as a name does.
• Prompt about the task, not the student. "A 5th-grade warm-up on fractions" gets you nearly everything "a warm-up for Marcus" would, with none of the exposure. Make it your default.
• Know what "approved" actually means. A tool that's genuinely safe for student work has, at minimum: a signed data-privacy agreement (DPA) with your district, a written statement that your prompts aren't used to train its models, documented FERPA compliance, and — for students under 13 — COPPA compliance. If a tool is listed on your state's student data privacy registry (many states run one), that's the fast check.
• Read the data terms of anything that hasn't cleared that bar. Know whether what you type is stored, for how long, and who can see it.

That's the checklist version. For the plain-language deep dive — what FERPA actually covers, how COPPA applies, and five "can I paste this?" scenarios worked through — see our guide to student data privacy and AI. Two questions from that territory come up often enough to have their own pages: how the two laws divide the work is in FERPA vs COPPA, and whether any ChatGPT account can be brought under FERPA is in Is ChatGPT FERPA compliant?

Do you need permission? The district-policy question

The awkward truth: as of RAND's 2023–24 survey, most districts hadn't told teachers anything — only 18% of principals said their school or district had provided guidance on AI use. A growing list of states now requires districts to adopt AI policies (our school AI policy template covers the mandates, and gives whoever inherits the job an adoptable draft), but plenty of teachers still decide alone, wondering whether the tool that saved them four hours of drafting is technically allowed.

A workable position until your district catches up: use AI freely for teacher work product that involves no student data (planning, materials, your own drafting); hold the line absolutely on student information; and ask for guidance in writing — it protects you, and enough teachers asking is what makes districts actually write the policy. If an administrator observes a lesson you planned with AI, the answer is the honest one: you drafted with AI and made every instructional decision yourself. Any principal who objects to that is objecting to the textbook's pacing guide too.

And the quieter worry — is using AI lazy? Planning from scratch was never the measure of good teaching; what happens in the room is. The teachers using six reclaimed hours on feedback and small-group time aren't cutting corners.

How to start without overhauling everything

1. Pick one recurring task you dislike — a warm-up, a rubric, a parent email.
2. Use AI for just that task, this week, with the prompt starters above.
3. Judge it honestly: did it save real time without dropping your standards?
4. Keep what passed. Drop what didn't. Add the next task only when the first one's boring.

The teachers who get the most from AI aren't the ones who use it for everything. They're the ones who found the three or four tasks where it genuinely delivers — usually planning, differentiation, and first-draft feedback — and gave the rest of their attention back to their students. If planning is where you'd start, Planning Partner drafts the whole lesson — plan, worksheet, slides, answer key — and leaves every decision that matters to you.