Using AI to Build Better Lesson Plans and Curriculum

Artificial intelligence does not replace that weight. It shifts it. Used well, AI lets you spend less time formatting and more time thinking about the person in front of you — the student, the resident, the client, the customer, the new hire, the parent in a workshop. That shift is what this post is about.

This is written for anyone who designs learning experiences — teachers, trainers, support professionals, healthcare educators, preceptors, residency directors, coaches, homeschool parents, workshop facilitators, onboarding leads, continuing education coordinators, and presenters. The tools and patterns that follow apply whether your audience is six-year-olds, adult learners returning to school, residents in a skilled nursing facility, clients in a day program, or software engineers in a bootcamp.

The core skill is the same across all of them: translating a goal into a sequence of experiences that help a specific person reach it. That is what AI can help you do more efficiently, more thoughtfully, and with more variation than most of us can manage alone at 9 p.m. on a Sunday.

Given a clear prompt, a capable model can draft most of the artifacts that make up a real curriculum. Not perfectly, but as a strong starting point you can edit in minutes instead of writing from scratch in hours.

The list is not exhaustive. If it involves writing, structuring, or formatting, AI can help with it.

The difference between a useless AI response and a genuinely helpful one is almost always the prompt. Vague prompts produce vague output. Specific prompts produce specific output. There are five parts worth including every single time.

1. Role. Tell the model what kind of expert to be. "You are an experienced middle school science teacher" produces different output than "You are a trauma-informed adult educator." The role shapes tone, vocabulary, and what the model considers default.

2. Audience. Describe the learner in concrete terms. Age range, prior knowledge, context, abilities, barriers, language level, cultural context. "A class of 11 to 13-year-olds, mixed reading levels, roughly 20 percent English language learners, some with IEPs for executive functioning" is a real audience. "Students" is not.

3. Objective. Say what the learner should be able to do by the end. Not what you will cover — what they will be able to do. "By the end of this lesson, the learner can identify the three branches of government and explain one check each branch places on the others in their own words."

4. Format. Specify the shape of the output. Lesson plan with these sections. Ten-slide deck outline. 90-minute workshop split into four blocks. One-page handout. The model will happily give you whatever shape you name.

5. Sources. Ask the model to cite where its information comes from and to flag anything it is not certain about. This single instruction changes the quality of the response — not because the model always cites perfectly, but because it slows down and is more careful with claims it cannot support.

Put these five parts together and you have a prompt that will almost always produce something useful.

Before the content, before the activities, before the assessment — describe the person you are teaching. This is where most lesson plans go wrong, and it is where AI can help you most if you let it.

When you prompt, give the model the details that actually matter for instructional design:

You do not need a research paper's worth of detail. A few sentences is enough. But those few sentences change everything about what the model will produce.

When curriculum designers talk about "learning styles," they most often mean the VARK model — Visual, Auditory, Reading/Writing, and Kinesthetic. It is a useful shorthand. It is also worth being honest about what the research says.

Beyond VARK, Howard Gardner's theory of multiple intelligences adds useful texture: logical-mathematical, interpersonal, intrapersonal, naturalistic, musical. Some people think best alone; some think best in dialogue. Some learn best in nature, some in silence, some while moving their hands.

Here is the honest part: pure single-style learners are rare. Most people are multimodal. They have preferences, not limits. The research evidence that teaching to a student's "preferred style" improves outcomes is thin — but the research that strong teaching hits multiple modalities is robust.

The practical move: do not try to sort your learners into boxes. Instead, build every lesson so that it reaches at least three modalities. Show it, say it, have them write about it, have them do it. Ask AI to generate lesson variations that add a missing modality to a plan you already have.

A good prompt sounds like this: "Take this lesson plan and add a kinesthetic activity that reinforces the same objective without replacing the existing visual and verbal instruction."

Person-centered is a term that started in disability services and healthcare but applies to every learning context. The idea is that the learner is the starting point, not the content.

In a person-centered frame, you ask:

This framework is especially powerful for adult learners, neurodivergent learners, learners with disabilities, learners with trauma histories, and learners in care settings. But it improves curriculum for everyone. A 9-year-old has as much right to be seen as a person as a 90-year-old does.

AI can help you operationalize this. Describe the learner in person-centered terms, and ask the model to generate a lesson that reflects them specifically. Not a generic "fifth-grade math lesson." A lesson for a specific kind of fifth-grader in a specific context.

An example prompt: "Design a 30-minute lesson on fractions for a 10-year-old who loves basketball, has ADHD, is on grade level in math but below grade level in reading, and does best when activities involve movement. Keep written text to a minimum. Use basketball examples. Include a one-minute movement break at the halfway point."

The resulting plan will be more useful than anything a generic template could produce.

Not every use case starts from a blank page. Often the most valuable work is taking something you already have and making it better. AI is excellent at this.

Paste in your existing plan and ask for specific enhancements. A few patterns that consistently work:

Each of these takes about 30 seconds to prompt and can add real pedagogical value that you would otherwise have to grind through on your own.

Models will sometimes invent sources. This is a known failure mode, not a character flaw. Your job is to catch it before it reaches a student.

Every time you ask AI for resources — articles, books, studies, videos, curricula — include these instructions:

Then fact-check. Paste titles into a search engine. Click through to the actual source. If the article does not exist, ask the model why it suggested it, and move on.

For high-stakes content — medical, legal, safety-related — do not rely on AI-generated summaries alone. Use them as orientation. Verify against primary sources before passing anything to a learner.

A few specific prompt patterns that work well:

Slide decks "Generate a 12-slide deck outline on [topic] for [audience]. Each slide should have a title, three to five bullet points, and a speaker note with a teaching tip. Include one discussion slide and one activity slide."

Facilitation scripts "Write the opening two minutes of this workshop as a script I can read aloud. Include a hook, an acknowledgment of why this topic matters, and a preview of what we will do together."

Worksheets "Create a one-page worksheet for this lesson. Use a mix of short-answer, sketch-or-label, and one open-ended reflection question. Include an answer key on a separate page."

Supply lists "Generate a supply list for a 90-minute hands-on workshop on [topic] for 15 participants. Assume a $75 total budget. Include quantities and note anything that needs to be prepped in advance."

Differentiated versions "Take this activity and produce three versions — one for learners who need more scaffolding, one for learners on target, and one that extends into a challenge for learners ready for more."

Each of these is a copy-paste starting point. Edit freely.

The question every curriculum designer now has to answer: how do my students use AI, and what do I teach them about it?

The honest answer: AI is a tool for thinking with, not thinking for you. That distinction is the whole lesson.

Some ways learners can use AI well:

Some ways learners misuse AI:

Teach learners to paste their own thinking in first and ask AI to respond to it. That single habit turns AI from a shortcut into a tutor. Teach them to ask the model to disagree with them. Teach them to verify, not trust.

If you work in healthcare, education, disability services, or any setting where you have protected information about the people you serve — do not paste that information into an AI tool. Names, birthdates, addresses, diagnoses, case numbers, medical details, and progress notes should stay out of the chat window.

When you need help with something client-specific, describe the situation generically. "A young adult in a day program who benefits from visual schedules and has difficulty with transitions" is enough for a helpful response. "Michael, who has been at our program since March, and was diagnosed with…" is a HIPAA problem waiting to happen.

The rule of thumb: if you would not say it in a crowded elevator, do not type it into a chat window. Anonymize everything. What remains will still produce useful output.

AI will not replace a good teacher, trainer, or curriculum designer. What it can do is take the mechanical parts of the work — the formatting, the boilerplate, the reshuffling — and give them back to you as drafts. That leaves you with more time for the parts that actually matter: reading the room, building trust, noticing when someone is lost, adjusting on the fly, and being present with the people in front of you.

Curriculum design has always been about seeing the learner clearly and then shaping the path to meet them. AI does not change that job. It just makes the path easier to draw.

Start small. Take a lesson you already teach. Give it to the model with five clear parts — role, audience, objective, format, sources — and ask for one enhancement. Edit the response. Use it next week. Notice what the model is good at and where it falls short.

That is where this work begins. The rest is practice.