For most teachers, yes — but the bar is lower than the marketing suggests and higher than the sceptics claim. At $14.95 a month, a paid plan pays for itself if it gives you back about 26 minutes of your time. Below that, the free tier is the right answer and nobody should sell you anything.
You make more than about two resources a month, or you want AI in front of students and marking what they write. The saving compounds; the price does not.
You generate a handful of resources a term, your board already funds a tool that covers planning, or you rewrite most of what the model produces.
The time you save goes somewhere other than more planning. Trials show use falling off over ten weeks — a subscription you stop opening is the expensive version.
A subscription is worth it when the value of the time it returns is larger than the price. That is the entire test. Move the sliders to where you honestly are, not where a demo day put you.
Three inputs, your estimates. This is arithmetic, not a promise — the trials below are the reality check on whether your minutes figure is believable.
1.3 hrs
120 min saved, less 42 min checking the output
$45.50
Plan costs $14.95 a month
26 min
About 2.6 resources a month
At these numbers the plan returns about $30.55 more than it costs you each month.
2,540 tokens per SBI4U lesson plan (median of 3 real generations, 27 October 2026), so the starter allowance covers roughly 19 resources a month.
We take 35% off your saving to account for checking the output before you use it. Change a slider and watch how little it takes to lose the case.
Minutes a month you need to get back before each plan pays for itself, at three different values for your own time. Below the line, you are working for the subscription.
| Plan | Price a month | If your hour is worth $25 | If your hour is worth $35 | If your hour is worth $50 |
|---|---|---|---|---|
| Starter | $14.95 | 36 min | 26 min | 18 min |
| Pro | $44.95 | 108 min | 77 min | 54 min |
So the whole question for a Starter subscriber is whether AI saves them more than half an hour a month. For the teacher who plans a full course, that is one lesson.
Two randomised trials and two surveys. The trials measure planning time directly; the surveys ask teachers to estimate. Read the trials first — the surveys are where the six-weeks-a-year headline comes from, and the median teacher reports a fraction of it.
Education Endowment Foundation / NFER
Lesson planning and preparation fell from 3 h 19 min to 2 h 30 min a week — about 49 minutes saved, a 24% reduction, with no measured loss of quality.
Primary schools only, and use of the tool declined over the ten-week trial. Fewer than one in five teachers felt it meaningfully changed their planning.
Read the sourceEducation Endowment Foundation / NFER (via Stanford SCALE)
Teachers using ChatGPT spent 69% of the planning time of the comparison group — 56.2 minutes a week against 81.5 minutes, about 25 minutes saved.
Both groups planned new lessons from scratch, which favours a tool that drafts from scratch. Time saved on adapting existing lessons is not measured here.
Read the sourceGallup × Walton Family Foundation, Teaching for Tomorrow
Teachers who use AI at least weekly estimate it saves them 5.9 hours a week — roughly six school weeks a year. Between 60% and 84% report time saved on each type of task.
These are self-reported estimates, and only about three in ten teachers use AI weekly. Very few (7% or less) say AI adds time.
Read the sourceNPR / Ipsos
62% of teachers use AI, and 63% of those who use it for work say it saves them two hours a week or less.
This is the median reality sitting underneath the six-weeks-a-year headline, and it is the number we would trust more.
Read the sourceGenerated content is a draft, and checking a draft is work. We take 35% off the saving in every calculation on this page to account for it — a lesson plan you would have spent forty minutes on, and now spend fifteen checking, has saved you twenty-five, not forty. If your review habit eats more than half, the honest verdict flips, and no tool fixes that but practice with the tool.
The same logic applies to curriculum alignment. If the output needs checking against a course expectation anyway, count that time. Ontario expectation codes attached to the resource exist to shorten that check, not to pretend it away.
Four ordinary situations where the arithmetic says keep your money. We would rather you decide with these in view than cancel after a term.
Two or three resources a month is roughly what a free tier covers. Paying for a subscription you open twice a term is the most common way teachers waste money on AI — the tool is fine, the subscription is the error.
If a licensed tool already covers planning time, a second one only earns its price by doing something the first cannot — putting AI in front of students, or marking what students actually write. Otherwise you are buying the same saving twice.
The arithmetic only holds if checking the output is a small share of the time saved. If you end up rewriting more than you would have written, the tool is costing you time and the answer is no, however good the demo looked.
Trials show tool use falling off over ten weeks. Time saved that goes straight back into more planning is not a saving — it is the same job with an extra subscription.
Do not trust this page, or any other. Measure one resource and do the subtraction yourself.
At
That figure comes from teachers who use AI at least weekly estimating their own savings at 5.9 hours a week — it is self-reported, not measured. The median teacher who uses AI reports two hours a week or less, and the randomised trials measure around 25 to 49 minutes a week on planning alone. The six weeks is a ceiling reported by enthusiasts, not a typical result.
In the 2026 Aila trial, quality was not measurably worse, but teachers were mixed on whether the tool changed their teaching, and the share using it fell over ten weeks. Save time first, then judge the output on your own students — not the other way round.
Often, yes. If free covers the volume you actually generate, the worth-it question is answered without spending anything. Only pay when you can point at a resource you would otherwise have made by hand and did not have time for.
You already have a tool your board funds, or you generate too little to need one. Both are ordinary, defensible reasons to keep the money, and they are the two cases this page is designed to surface rather than talk you out of.
Generate one unit's worth of resources, time the difference, and decide with a number instead of an opinion. No card, and nothing to cancel if the answer is no.