How to check an AI lesson plan before you teach it
Auto Classmate
11 min read
To check an AI lesson plan before you teach it, test four things: the content is accurate, the activities match the standard the plan names, the thinking level is high enough, and the timing and supports fit your students. Work every problem, read the standard’s actual text, and rehearse the hardest minute. Until then, it is a draft.
That check matters more in 2026 because most teachers are doing it alone. A Gallup and Walton Family Foundation study published in May 2026 found that six in 10 U.S. K-12 teachers use AI for their work, yet only 18% receive any formal guidance on how to use it. For preparing to teach, including planning lessons, 45% reported no guidance at all.
This guide gives you a repeatable review built around one running example: a 7th grade math lesson on the constant of proportionality, taught in a 47-minute period. Swap in your own grade and subject. The checks stay the same whether you teach kindergarten phonics or AP Chemistry.
What does “correct” mean for an AI lesson plan?
A correct AI lesson plan passes four tests, and only one of them is about facts. A correct lesson plan is one that is accurate, aligned to its standard, rigorous enough for the learning goal, and runnable with the students in front of you. AI tools are good at producing the shape of a plan: objective, hook, guided practice, independent practice, exit ticket. The shape is rarely the problem.
| Test | The question you ask | What failure looks like |
|---|---|---|
| Accurate | Is every fact, example, and answer key right? | A worked example with an arithmetic slip, a wrong date, a key that disagrees with its own problem |
| Aligned | Do the activities teach the standard the plan names? | The plan cites 7.RP.A.2, but every task is unit-rate practice |
| Rigorous | Do students analyze, justify, or create during class time? | Tasks labeled “analyze” that only ask students to restate a definition |
| Runnable | Does it fit your bell schedule, materials, and roster? | 58 minutes of activities in a 47-minute period, with no named supports |
Accuracy is the test most teachers reach for first. Alignment and rigor fail more quietly, because a plan can be factually clean and still teach part of a standard at the wrong depth. That is why this checklist gives each test its own step.
Why can’t you trust an AI lesson plan as written?
You can’t trust an AI lesson plan as written because it is fluent text generated from patterns, not a judgment about your standard or your students. The research on how to evaluate AI-generated lesson plans points to predictable weak spots, and knowing them tells you where to look.
The U.S. Department of Education’s 2023 report on AI in teaching and learning warned that AI can produce “information that appears authentic, but actually is inaccurate.” Its central recommendation is keeping humans in the loop, with teachers remaining in charge of major instructional decisions. A review routine is how that recommendation shows up on a Tuesday night.
So are AI lesson plans accurate? Facts are only part of the answer. The larger documented gap is thinking level. A 2025 peer-reviewed study by Torrey Trust, Robert Maloy, and colleagues at UMass Amherst, published in CITE Journal, analyzed 310 lesson plans that three popular chatbots generated for Massachusetts 8th grade civics standards. Of 2,230 activities, 45% asked only for recall and just 10% asked for higher-order thinking.
Two more findings from that study shape the steps below. The higher-order tasks that did appear were concentrated in homework and extension sections, not class time. And the researchers coded each activity by what students would actually do, because many tasks labeled “analyze” only asked students to explain. Adding “highly interactive” to the prompt produced only minor changes. The authors’ advice to teachers is plain: choose “not to adopt any AI-generated lesson plan exactly as it is written.”
How do you check an AI lesson plan, step by step?
Run these seven checks in order. The first three catch errors on the page. The last four catch problems that only show up in a room with real students. The running example is an AI-drafted plan titled “Finding the constant of proportionality,” tagged to Common Core 7.RP.A.2, for a 47-minute 7th grade math period.
- Read the standard itself, not the plan’s summary of it. Open the official text and compare it with the activities. Common Core standard 7.RP.A.2.b asks students to identify the constant of proportionality “in tables, graphs, equations, diagrams, and verbal descriptions.” If every task in the plan uses tables, the lesson covers one representation, not the standard. Narrow the objective or add a graph task. The free Standards Unpacker turns one standard into I-can targets, misconceptions, and evidence.
- Work every problem and check every key. Solve each example and practice item yourself, on paper, with the answer key covered. A table whose y-to-x ratios run 3, 3, 3.5 is not proportional, whatever the key says. Keep it as a deliberate non-example if you like, but label it that way. In science, social studies, and ELA, check every fact, date, and quotation against your adopted curriculum or the original text.
- Label each activity by what students actually do. Ignore the verb in the heading. Depth of knowledge (DOK) is Norman Webb’s four-level scale for the thinking a task demands, from recall (Level 1) to extended thinking (Level 4). In his 2002 depth-of-knowledge descriptors, Webb notes that verbs like “describe” and “explain” can sit at different levels depending on what students are describing, and that asking students to explain their thinking usually reaches Level 3.
The classroom move: change “Find k for each table” (Level 1 or 2) to “Two of these four tables are proportional. Which two, and how do you know? Convince your partner.” Same content, same minutes, and now students have to reason and defend a claim during class instead of in an optional extension. - Do the arithmetic on the clock. Add up every timed segment and compare the total with your real bell schedule, not a generic 60-minute period. Count what drafts leave out: settling in, passing out materials, the transition from partners back to seats. If the draft totals 58 minutes for a 47-minute period, decide tonight what goes. Cut the second practice set before you cut the check for understanding.
- Check it against the students in your room. Name supports by group: a sentence frame (“The constant of proportionality is ___ because ___.”) for students at early English proficiency levels, a partly completed table for students whose IEPs reduce written load, and a tricky non-example for students who finish early. If the draft’s supports are generic, the free Differentiation Planner generates ELL, IEP, gifted, and scaffolding moves tied to your own lesson.
The Department of Education report makes the same point with a math example: three students make an identical error for three different reasons, and only the teacher has the context to respond to each one. No draft can see that. You can. - Place a check for understanding mid-lesson and decide what it triggers. In the UMass study, most plans placed assessment at the end. Put a hinge question at minute 12 instead: project one table and have students show one finger for proportional, two for not. Decide in advance what a wrong answer triggers. If more than a handful of students miss it, reteach with a graph (a straight line through the origin) before anyone starts independent practice.
- Rehearse the hardest minute. Read the plan aloud the way your students will hear it and find the moment most likely to stall: a three-step direction packed into one sentence, a transition with no signal. The free Lesson & Activity Forecast runs a version of this rehearsal for you. Paste the plan with student names removed, and it predicts student reactions, friction points, and fixes, and does the arithmetic on your timings.
Which parts of an AI lesson plan should you check first?
If you only have five minutes, check the answer key, the standard match, and the clock, in that order. Those three are the fastest to verify and the most disruptive to discover in front of 30 students. The table below covers every section, so you can work down it when you have more time.
| Plan section | What to look for | Fast fix |
|---|---|---|
| Objective and standard code | A missing code, the wrong grade, or activities that teach a different standard | Paste the official standard text above the objective and edit until they match |
| Content and examples | Wrong facts, invented quotations, arithmetic slips | Verify against your curriculum or the original source |
| Answer keys | Keys that disagree with the problems | Solve every item with the key covered |
| Activities | Recall tasks labeled “analyze” or “evaluate” | Rewrite one in-class task so students must justify an answer (DOK 3) |
| Timing | Segments that exceed your period or skip transitions | Add the minutes, then cut the second practice set |
| Supports | Generic “differentiate as needed” lines | Name one scaffold for each student group |
| Assessment | One exit ticket and no mid-lesson check | Add a hinge question between minutes 10 and 15 |
The checks hold across grades, but the likeliest failure point shifts. In K-2, read the directions aloud and count the steps; a four-step direction is the weak point for a six-year-old. In high school science, check units, safety notes, and whether the lab runs with the equipment you actually have. In ELA, confirm that every quoted line exists in the text students will read.
Does checking an AI lesson plan cancel out the time it saves?
No. Reviewing an AI lesson plan takes minutes when you check for the four failures instead of rewriting the plan from scratch. The time saved comes from not building structure you already know how to build. It does not come from skipping the review.
A better way to think about it: the draft moves your time from assembling a plan to judging one. Judging is the part that needs a teacher. A tool can fill 47 minutes with activities; it cannot know that your fourth period loses five of those minutes to a late lunch bell unless someone tells it.
What makes an AI lesson plan easier to check?
A plan that starts from your context leaves less to fix. A general chatbot is stateless: it does not know your period length, your roster’s needs, or the exact wording of your standards unless you paste them in every time, and it rebuilds the lesson structure from zero with each request. Every missing detail becomes one more thing you check by hand.
That is the gap Auto Classmate’s AI lesson planning is built to close. It learns a teacher’s class profile, standards, and teaching style, so plans start from your room rather than an average one. Ten standards frameworks, including the Common Core, ship built in and unpacked into learning targets, and reflection memory keeps a record of what happened when you taught a lesson. More than 65,000 teachers in 120+ countries use it (our figure).
Privacy is part of what makes a plan checkable, because you should never have to paste identifiable student details to get a useful draft. Auto Classmate’s class profiles hold group-level information only, no field asks for student names, and lesson data is never used to train outside AI models. The platform is designed with FERPA, the federal student education records law, in mind. None of this removes the check. It shortens it.
Frequently asked questions
Can I trust an AI lesson plan if it lists a standard code?
A standard code on an AI lesson plan is a claim, not proof of alignment. Open the official standard text and compare it with what students actually do in each activity. A plan can cite a real, correct code and still teach only part of it, such as practicing a skill with tables when the standard also requires graphs, equations, and verbal descriptions. Edit the objective or the tasks until the two match.
What should I check if I only have five minutes?
Check three things in this order. First, solve the practice problems with the answer key covered, because a wrong key derails a lesson fastest. Second, read the standard’s official text against the activities. Third, add up the minutes and compare the total with your real period length. Rigor and differentiation matter just as much, but those three errors are the quickest to catch before class.
How do I make an AI lesson plan more rigorous?
Rewrite at least one in-class task so students must justify an answer, not just produce one. In Norman Webb’s depth-of-knowledge descriptors, asking students to explain their thinking usually reaches Level 3. Then check the extension section. A 2025 UMass Amherst study of 310 AI-generated plans found higher-order tasks clustered in homework and extensions, so move the strongest one into class time.
Is it safe to put student information into an AI lesson planner?
Never enter student names or identifiable details into an AI tool unless your district has approved it for that use. Describe your class at the group level instead: how many students are English learners, which reading levels you teach, which accommodations appear across IEPs. Check your district’s AI and data privacy policy first, since student education records are protected under the federal FERPA law.
Should I tell my principal or coach that I used AI to plan?
Follow your school’s expectations, and ask if there are none. Gallup’s 2026 teacher survey found that fewer than one in 10 teachers get formal guidance on any specific AI task, so many schools simply have not said. Keeping brief review notes (what you changed and why) makes the conversation easy, because it shows the professional judgment behind the final plan.
How can instructional coaches help teachers review AI lesson plans?
Use the seven checks as a shared protocol in a PLC or planning meeting. Have each teacher bring one AI-drafted plan, label every activity’s DOK level together, and compare those labels with the headings the AI wrote. The gaps between the two make a concrete coaching conversation. Department heads can also standardize a minute-12 check for understanding across sections of the same course.
Try it on tomorrow’s lesson. Run the first six checks yourself, then paste the plan into the free Lesson & Activity Forecast to rehearse the hardest minute before your students do. No account needed, three forecasts a day.