Ask almost any AI tool for a lesson “aligned to” a standard and you will get one back with the code printed at the top. That is labelling. Whether the lesson is aligned is a separate question, and the code at the top tells you nothing about it.
Labelling versus alignment
A standard is a claim about what a student can do, stated with a specific verb at a specific depth. Alignment is the property that the work students actually produce would be evidence for that claim. Those two things come apart constantly, and the failure is almost always in the same direction: the lesson lands one level below the standard.
- The standard says analyse how an author develops a theme; the exit ticket asks students to identify the theme. Same topic, lower demand, no evidence for the standard.
- The standard says construct an explanation for a phenomenon; the task asks students to label a diagram. The diagram is fine — it is just not the claim.
- The standard says compare two accounts; the packet has students summarise each one separately and never compare them.
Each of those lessons could carry the standard code truthfully in the sense that the topic matches. None of them produce the evidence.
Why generation drifts downward
Two structural reasons, both fixable. First, a model asked for “a lesson on X” optimises for a coherent, teachable activity, and lower-demand activities are more coherent to write. Second — and this is the bigger one — most tools generate each activity in isolation, with no idea what the rest of the unit already did. Without that, every day has to be self-contained, and a self-contained day cannot ask students to build on Tuesday.
A two-minute check for any generated lesson
- Find the standard's verb — analyse, construct, compare, evaluate. Circle it.
- Find the single piece of work a student hands you at the end of the lesson.
- Ask whether that artifact could be evidence for the verb. Not the topic — the verb.
- If it could not, the lesson is labelled, not aligned. Usually the fix is one level of demand in the final task, not a new lesson.
How TeacherHero handles this
Two design decisions, both deliberate, and neither of them is “we solved alignment.”
- The final product is required before anything generates. You must state what students will produce before a single section can be written, because every section is written backwards from that answer. A blank one is not allowed to proceed — that is enforced in code, not suggested in the UI.
- Alignment is a check you read, not a badge you receive. A lesson can be checked against its own learning target, producing findings you can act on. The result is cleared automatically whenever the lesson's context changes, because a check is only valid for the version it was computed on. A stale green light is worse than none.
You still make the call. A tool that tells you a lesson is aligned has replaced your judgement with its own; a tool that shows you where it drifted has given you something to act on.