Why ChatGPT Writes Generic Lesson Plans (and the One Document That Fixes It)

A teacher asks ChatGPT for a photosynthesis lesson and gets back a tidy 60-minute lecture plan for a class that does not exist. Her periods run 47 minutes. Her readers span 4th to 9th grade level. Nothing aligns to her standard, and the whole thing assumes students who sit quietly through twenty minutes of direct instruction. So she rewrites it, concludes AI planning is overrated, and either quits or pays $20 a month for a teacher AI tool. There is a third option, and it costs one page of writing.

Why does ChatGPT give me generic lesson plans?

Because a language model fills every gap you leave with the statistical average, and the average classroom is not yours. These tools generate the most plausible response to exactly what is in the conversation. They do not know your state standards, your bell schedule, your pacing guide, or your readers unless you tell them, so an unspecified request gets the average answer: a 60-minute lesson, on-grade texts, lecture plus worksheet. The output is not wrong so much as unspecified. Specify the gaps once and they stay specified for the whole conversation.

What is a Master Context Document?

One page that describes your real classroom, written once and pasted at the top of any AI chat before you ask for anything. The strong version carries about ten facts: your standards framework with the current unit codes copied from your own district guide, your pacing anchor, your true period length minus transition time, your reading-level spread described as a distribution, your pedagogy preferences, your materials, your assessment style, your district AI policy, and the plan format you want back. After that single paste, every lesson request is about three lines long: the topic, anything different about today, and the output format. The context carries the other 90 percent of the specification.

This is, mechanically, the same thing the $15 to $20 per month teacher AI tools do behind the scenes: they store a profile and inject it into prompts. A document you own does the same job on the free tiers, and it never goes stale when the models update, because pasting a page of text at the top of a chat is the one interface every chatbot shares.

I need to plan next week's lessons tonight. What do I actually do?

Twenty minutes, five moves, one chat. If that heading is roughly what you just typed into a chatbot on a Sunday evening, here is the whole workflow:

1. Open a fresh chat and paste your Master Context Document. Ten seconds if it is written; thirty minutes once if it is not, and that investment pays back every week for the rest of your career.
2. Orient the chat: paste last week's carry-forward summary plus one honest line about what actually happened and what the exit tickets showed.
3. Request the week: ask for the next lessons from your unit map in your standard format, each ending with its own carry-forward summary. Note anything odd, like an assembly schedule on Wednesday.
4. Refine with one-line fixes: swap a flat hook, resize an activity, retier a text. Two or three lines is typical.
5. Verify before you teach: circle every date, fact, quote, and answer-key item and check them. Five minutes, never optional.

Then paste the plans into your binder or LMS and close the laptop. That is the entire Sunday routine the system is built around.

Can free AI plan a whole unit without losing the thread?

Yes, if every lesson hands a summary forward to the next one. The reason unit planning collapses is that lesson 4 has no idea what lessons 1 through 3 taught, so it reteaches vocabulary and contradicts your examples. The fix is a habit: end every generated lesson with a five-to-eight line carry-forward summary covering the objective met, the vocabulary as taught, the anchor example used, and the misconceptions your exit tickets caught. Paste that summary at the top of the next lesson request. Edit it first to match what actually happened in class, because the chain should describe reality, not the plan. Continuity, solved, on a free tier.

What is safe to paste into ChatGPT as a teacher?

Classroom-level information goes in. Student-level information never does. Describe your class as a distribution: "reading levels span 4th to 9th grade, about a third below grade level." Never as individuals: no names, no initials, no gradebook exports, no IEP text, no student work with identifying details. If a paste would let anyone identify a specific student, it does not go into a consumer chatbot, free tier or paid. Held consistently, that one line keeps you on the right side of FERPA caution and your district policy, and it costs nothing in output quality, because planning needs patterns, not identities.

One more non-negotiable: verify every specific fact, date, quote, and answer key before it reaches students. These models are confidently wrong in ways designed to slip past a tired skim. Circle the checkable claims and check them. It takes about five minutes per plan and it is not optional.

Do I need a paid AI subscription to plan lessons?

No. A planning session is one context paste plus a handful of requests, which fits comfortably inside the free tiers of ChatGPT, Claude, and Gemini. If one free tier rate-limits you, the system is portable: open another tool and paste the same document. That portability is the point. You are not learning a product that can raise prices or shut down. You are learning how the technology works, on paper you control.

Want the complete system?

The full method is packaged as The Teaching Context System: Make Free AI Plan Like It Knows Your Classroom: a 50-page guide, fill-in Master Context templates for K5, 6 to 12, and higher ed with a completed example beside every blank, the unit-chain worksheet, a printable FERPA-safe checklist for your desk, and 25 ready prompts that inherit your context automatically. One-time purchase, built for the free tiers.

→ Get The Teaching Context System ($39)

Teaching at the college level? The policy side of this problem, syllabus AI statements, assignment design, and integrity conversations, is covered in The Syllabus AI Policy Kit.

Frequently asked questions

Does this work with Claude and Gemini or just ChatGPT?
All three, and whatever ships next. The method is deliberately tool-agnostic: a context document pasted at the top of a chat works in every chatbot ever made, which is why it does not expire when products update.

Will AI invent standards codes?
Yes, if you let it. Models will produce plausible-looking codes that are wrong or do not exist. Always paste your current unit codes from your own district documents and instruct the AI never to supply codes from memory.

I want AI to differentiate a lesson for my below-grade-level readers. Can it?
Yes, and this is where the context document earns its keep: once your reading-level spread is loaded, you can ask for every text at two levels, directions in numbered steps, and a three-tier version of the core activity with the same objective in each tier. Same goal, different route.

Does AI replace teacher judgment?
No. AI drafts; you decide. The system treats AI as a fast, occasionally wrong assistant: it inherits your constraints, you verify its facts, and your exit tickets, not the model, decide what gets retaught tomorrow.

How long does setup take?
About 30 minutes once to fill in the Master Context Document, using facts you already know. After that, teachers typically plan a week of aligned lessons in about 20 minutes, including the verification pass.