How to check AI generated lesson plan drafts before class: a practical checklist

To check AI generated lesson plan drafts before class, confirm that the plan matches your learning goal and grade level, verify every factual claim and standard reference against a source you trust, and test any rubric by scoring real or sample student work. Then make the judgments only you can make, such as what your students already know and what your school allows. Practice on deliberately flawed material first, so the habit is ready when a real plan arrives.

Where AI lesson plans and rubrics go wrong

AI drafts rarely look broken. Headings are tidy, the tone is confident, and the timing table adds up on paper. That polish is why the failures are easy to miss on a quick skim. Most problems fall into a few repeatable patterns, and knowing them tells you where to look.

Errors that look finished

Invented or misattributed facts are the most common. A date is slightly off, a quotation belongs to someone else, or a statistic appears with no source. Activities can also drift from the objective: the stated goal is analyzing a primary source, but the task is a worksheet of definitions. Watch for examples that are too advanced for the grade or so simple that students learn nothing. Standard codes are another weak point, because a code that looks right may point to a different grade band.

Rubrics that score the wrong thing

Rubric problems are quieter. Criteria overlap, so a student earns credit twice for one thing. Descriptors such as 'good' or 'clear' name no evidence a grader could point to. Point totals may reward length over accuracy, so a long, wrong answer beats a short, correct one. The scale may also fail to add up to the grade you intended. Each of these looks reasonable until someone actually scores a paper.

A five-pass check for any AI lesson plan

Work through the plan in a fixed order. Each pass depends on the one before it, so do not judge the activities until you know the goal is right.

1. Learning goal. Does the objective name something students can show? 'Understand photosynthesis' is hard to observe. 'Label the inputs and outputs of photosynthesis on a diagram' can be checked.
2. Activity alignment. Does every activity serve that objective? Mark any task that practices a different skill, then cut it or move it.
3. Facts and references. Verify dates, names, quotations, and standard codes against a source you trust, such as a textbook, a standards document, or a reputable reference.
4. Level and difficulty. Read the text as a student would. Check reading level, vocabulary, and whether the examples fit your grade.
5. Evidence of learning. Is there a way to see whether students learned it, such as an exit question or a short task scored against the objective?

Here is a pass 3 failure. The draft says: 'Marie Curie won the Nobel Prize in Physics in 1911.' Her physics prize was in 1903, and 1911 was her chemistry prize. Do not fix this from memory. Open a reliable biography or the official Nobel Prize site, confirm each year against each prize, and note the source beside the line. If the source disagrees with the draft, the source wins. Fix the sentence and record the change.

Testing a grading rubric before you hand it out

Reading a rubric is not the same as using it. Test it on real or sample work before students see it.

Pick three responses: one strong, one middling, one weak. If you have no student work, write them yourself or ask a colleague. Score each one using only the rubric, without relying on your own expectations. Keep notes on three things: where you hesitated, where two criteria gave the same credit for the same thing, and where the total felt wrong. A rubric that gives a strong answer and a weak answer the same total has failed, even if every line reads well.

Then revise from your notes:

• Merge or sharpen criteria that overlap, so each one measures a single skill.
• Replace vague descriptors with observable evidence. 'Uses specific evidence from the text' works better than 'good analysis'.
• Check that accuracy counts for more than length, and that the top score requires correct content.
• Confirm the points add up to the grade scale you plan to use.

Try the revised rubric on the same three samples. If the scores now separate them clearly and you agree with the result, the rubric is ready. If not, revise again. Three samples is a small test, but it can reveal scoring problems that a quick read misses.

The checks only you can make

An AI tool cannot know your room. Some judgments stay with you, and it helps to write them down.

• Fit with prior learning. Has the class covered the prerequisite ideas? A plan that assumes an untaught skill will fail regardless of its quality.
• School policy. Does your school allow AI use for this task? Does the plan ask students to share personal information with a tool? Check your district's rules before using AI-generated material with students.
• Representation. Do the examples reflect your students' lives without stereotypes? AI drafts often default to narrow or outdated scenarios.
• Timing. Does the schedule fit your period length, including transitions and cleanup? Estimated minutes tend to be optimistic.

Keep a short personal checklist with these four items and reuse it. Also note which AI-generated parts you changed, for example: 'rewrote Activity 2, replaced two examples, corrected the Nobel date.' This keeps the decision visible to you later and to colleagues who may reuse the plan.

Practice the checks before the next planning session

Checking gets easier with repetition, so practice before a real plan depends on it. Toula Studio has three practice tools that cover different parts of this routine. They are practice tools, not checkers. They do not review your plans for you, and they do not replace checking against your own sources.

• Verification Lab. Review a deliberately flawed lesson plan or grading rubric, then get coaching feedback on the checks you made. It is a safe place to miss an error and learn from it.
• Prompt Repair Studio. Improve the prompt behind a plan. It scores clarity, context, constraints, and verification, and gives you a rewritten prompt. A tighter prompt usually means less fixing afterward.
• Quantum Tutor. Ask a clear question about a specific claim you are unsure of. You get a clearer restatement of your question, why the change helps, three to five steps for checking the answer, and a responsible-use habit.

To start, try the Verification Lab on a flawed sample rubric on the Toula Studio website. You can also open the sample lesson there without an account. Toula Studio offers a free Explorer plan if you want to keep practicing.

Frequently asked questions

How long should checking an AI generated lesson plan take?

There is no fixed time, because it depends on the plan's length and how much it relies on facts. A quick pass on the learning goal and alignment can happen in one sitting. Testing a rubric takes longer because you have to score samples. Spend the most time on parts that carry grades or that teach factual content.

Should I tell students when I used AI to build a rubric or lesson?

That decision depends on your school's policy and your own judgment. Many teachers add a short note saying which parts were drafted with AI and how they were checked. This also teaches students that claims need verification. Check your district's guidance first.

What do I do if the AI plan is mostly good but has one or two errors?

Fix the errors you have verified, record each change, and keep the parts that match your goal. A mostly good draft does not need to be discarded. Do not teach it, however, until every fact and standard reference has been checked against a source you trust.