AI for Teachers.
AI will not replace the thing that makes you a teacher — the relationship, the read of a room, the moment you know exactly what a struggling student needs. It will absolutely replace the Sunday evening you spend building a lesson plan from scratch, differentiating a worksheet into three reading levels, and drafting the same feedback comment for the thirtieth time. This handbook is the practical middle ground: where AI genuinely helps teaching, the three rules that keep it safe for students, and five workflows you can use this week — no technical background required.
Where AI actually helps — and where it doesn't
The single most useful mental model: treat AI like a fast, tireless teaching assistant with no license, no knowledge of your specific students, and occasional confident lying. You'd happily hand that assistant a lesson-plan draft or a worksheet to differentiate. You would never let it assign final grades unchecked, and you'd never let it near identifiable student records.
| Task | Verdict | Why |
|---|---|---|
| Lesson plans, activities, discussion questions | Excellent | Blank-page problem solved; you shape and correct instead of compose |
| Differentiating materials — reading levels, scaffolds, extensions | Excellent | Same content, many versions, in minutes instead of hours |
| First-draft feedback & rubrics (anonymized) | Strong | Specific, actionable comments you then edit and own |
| Quizzes, worksheets, practice problems | Strong | Fast generation; you check every answer key (Rule 2) |
| Parent emails & newsletters (no student specifics) | Strong | Draft the tone and structure; you add the real facts |
| Facts, dates, quotes, answer keys — unverified | Never unverified | Models state wrong facts and wrong answers confidently — see Rule 2 |
| Final grades, or anything with named student data | Never | Your judgment and students' privacy — see Rules 1 & 3 |
The three non-negotiable rules
Rule 1 — Identifiable student data never goes into a consumer AI tool.
Free and consumer AI products may retain what you type and use it to improve future models. A named student's essay, grades, IEP or 504 details, behavior notes, or family information are protected education records — under FERPA in the US and similar laws elsewhere — and once pasted, you cannot un-paste them. The discipline: use school-approved education- or enterprise-tier tools with contractual no-training and retention commitments, and remove student names and identifying details even then ("a 7th grader writing below grade level" — not the name). If your district hasn't approved a tool, the safe default is: don't put student data in it.
Rule 2 — Verify every fact and every answer. Every one.
Language models generate plausible text, not verified facts — they will state a wrong date, misattribute a quote, or produce an answer key with confident errors, especially in math and at the edges of a topic. You are the subject-matter expert; the AI is a fast drafter. Every fact you'll teach, every worked example, every answer key gets checked before it reaches a student, because a confidently-wrong worksheet teaches the wrong thing to thirty kids at once.
Rule 3 — You are still the teacher.
Pedagogy, relationships, and judgment do not delegate to software. AI can draft a lesson, but only you know that this class needs the hook re-done, that this student needs a different entry point, that today isn't the day for group work. AI can draft feedback, but the grade — and the human read of a student's growth — is yours. The time you save is real, and the point of saving it is to spend more of yourself on the parts of teaching that are irreducibly human: not less teaching, but more of the teaching that matters.
Five workflows you can use this week
Each recipe: what to give the AI, what to ask, and what you must check by hand. The prompts are starting points — adjust for your grade, subject, and students, and keep the privacy discipline.
1 · Build (and improve) a lesson plan
- Give grade, subject, objective, time, and constraints:
2 · Differentiate a text or task into levels
- Provide the base material and the range you need:
3 · First-draft feedback on anonymized work
- Remove the student's name and any identifying details first (Rule 1).
- Give the rubric and ask for actionable feedback, not a grade:
4 · Generate a quiz or practice set — then check the key
- Specify topic, format, difficulty spread, and length:
5 · Draft parent communication
- Give the situation and tone — no student names or sensitive specifics in a consumer tool:
The judgment exercise: spot the danger
Three scenarios from real classroom patterns. Pick what you'd do — the point is calibrating when the rules bite.
1. An AI detector flags a student's essay as "98% AI-generated." The writing did seem more polished than usual. Do you accuse them?
AI detectors are notoriously unreliable and disproportionately false-flag non-native English writers and strong students. A detector score is not evidence. The durable move is process-based: ask about their writing, look at drafts and revision history, and design assignments (in-class writing, outlines, oral defense) that make process visible — not an arms race you'll lose.
2. You want AI feedback on a struggling student's essay, and it's fastest to just paste it into a free chatbot with their name on top. Do you?
Rule 1. A consumer tool may retain and train on the essay; deleting the chat doesn't undo retention on the provider's side, and student work is protected under FERPA. Anonymize first and use a district-approved tool — the feedback is just as useful without the name.
3. The AI generates a beautiful history worksheet with a full answer key. It's due to be copied for tomorrow. Use it as-is?
Rule 2. "Looks thorough" is the model's specialty and says nothing about whether the dates are right or the answer key is correct — models confidently misstate historical facts and answers. A worksheet with a wrong key or a fabricated fact teaches the error to the whole class. You're the expert; verify before you copy.
Choosing tools: the questions that matter
You don't need to understand the technology to procure it well. You (or your district) need answers, in writing, to five questions:
| Question | Answer you want |
|---|---|
| Is student or teacher data used to train your models? | No, contractually — not "you can opt out somewhere in settings" |
| Is it compliant with student-privacy law (FERPA, COPPA, state laws)? | Yes, with a signed data-privacy agreement |
| How long is data retained, and can it be deleted? | Defined, short, deletable on request |
| Security posture? | SOC 2 Type II or equivalent, encryption in transit and at rest |
| Admin controls and visibility for the school? | Yes — the district should be able to manage and audit use |
General-purpose enterprise AI (Claude, ChatGPT's business/edu tiers) covers the recipes here for your prep work. Education-specific platforms add curriculum alignment, LMS integration, and student-facing controls — evaluate those through your district's privacy review. And if your school has an AI policy, it precedes everything here; if it doesn't, these three rules are a reasonable seed for one.
Quick answers
Should students be allowed to use AI at all?
The realistic answer is that they already can and will, so the durable move is to teach responsible, disclosed use as a thinking tool — and to design assignments that value process and personal insight AI can't fake — rather than to ban and police. Policies vary; follow your school's, and make expectations explicit per assignment.
Can I trust AI to do the math in a worksheet?
No — re-work every answer key yourself, especially for math. Models make confident arithmetic and logic errors. Use AI to draft the questions; use your own work to verify the answers.
Does using AI to plan lessons make me a worse teacher?
Only if you ship its output unexamined. Used well, it removes the blank-page grind so you spend your energy on pedagogy, differentiation, and students — the parts that actually make you a good teacher.
Where should a skeptical teacher start?
Recipe 2 — differentiating a text you already use into reading levels. Zero privacy risk, immediately useful, and it shows the genuine strength (fast transformation of your own material) without touching grades or student data.
Related on Vibe Engines
Explore the topic
See this alongside everything else on the same subject — handbooks, system designs, challenges and tools, in one place.
More Handbooks
- The Prompting HandbookA friendly, hands-on field guide for everyday humans — learn the CRISP framework, spot bad prompts, practice with real recipes, play a drag-and-drop game, and test yourself with a quiz. No code required.Read →
- The Agentic AI Interview HandbookTwenty topics every senior AI engineer should be able to reason about live — from eval pipelines to reliability patterns for generative systems.Read →
- The Senior AI Engineer Interview Handbook60 questions across architecture, production incidents, agentic systems, RAG, evals, cost, safety, and leadership — what staff-level AI interviewers actually probe for.Read →
- 50 Angular Interview QuestionsA visual handbook covering components, change detection, RxJS, signals, routing, forms, performance, and testing — what interviewers actually probe for in senior Angular roles.Read →
- 50 Python Interview QuestionsFundamentals to advanced: data structures, OOP, iterators & generators, the GIL, asyncio, memory, testing, and the standard library — a visual walk through everything a Python interview touches.Read →
- 51 LLM Evals Interview QuestionsGolden sets, LLM-as-judge, regression testing, offline vs online evals, RAG evals, agent evals, red-teaming, and observability — demystified for interviews and production.Read →
Explore more from Vibe Engines
Get the next one in your inbox.
New handbooks, system-design walkthroughs, and tools — straight to your inbox. No spam, unsubscribe anytime.