Not a teacher, but I coach writing and I work with a lot of educators who are trying to figure out the same thing. The AI teacher tools landscape has exploded in the last eighteen months and most of it seems designed to demo well rather than survive actual classroom use.
The common thread I keep hearing from teachers I talk to: the tools that look impressive in a product demo collapse when you try to adapt them to a specific class, a specific curriculum, a specific group of students. The AI doesn’t know your students. It doesn’t know what you covered three weeks ago. It produces generically sound material that requires significant reworking to be actually useful.
The tools that seem to hold up are the narrow ones. A tool that does one thing well, like generating quiz questions from a passage you provide, or creating rubric variations, or differentiating a reading for multiple levels, beats a platform that claims to do everything.
What are educators actually getting sustained use out of, beyond the novelty period?
The narrow tool insight tracks with what I see in professional development contexts generally. People get sustained use out of tools that fit cleanly into an existing workflow and don’t require learning a new way of working. Tools that replace the workflow rather than extending it see high churn after the novelty period, just as you described.
Rubric generation has been useful for me as a starting point, same principle. The AI version is never right as-is but it surfaces criteria I might not have articulated explicitly and gives me something to react to. Generating against is faster than generating from scratch. The output is a draft, not a deliverable.
The quiz generation from provided text has been genuinely useful. I paste in a reading and get question options that I then edit and curate. The tool handles the mechanical work of generating options, I handle the pedagogical judgment of what’s actually worth asking. That division of labor works. The tools that try to handle the pedagogical judgment too are where it breaks down.
The planning time return is worth thinking about carefully. If learning a tool takes four hours and saves you twenty minutes a week, the math doesn’t work until you’ve been using it for six months. The narrow tools that do one clear thing tend to have faster payback because you learn them quickly and apply them immediately. The platform tools have a much longer payback horizon.
from a student side, i actually like when teachers use differentiated reading levels. the easier version of a complex text helps me understand the hard version better. so whatever tool is doing that seems to be working from where i’m sitting