On this page
Many teachers now ask an AI chatbot for a lesson outline, a worksheet or ten quiz questions. It saves time, but an AI draft can contain a wrong number, an outdated idea or a question pitched at the wrong level, and students will take it as correct because it came from you. This post covers what the strongest trial so far found about time and quality, and gives a checklist to run on every AI draft before it reaches a class.
Does AI actually save teachers time?
The strongest evidence so far comes from England. The Education Endowment Foundation funded a randomised trial, evaluated independently by NFER, in which 68 secondary schools and 259 science teachers were split into two groups. One group used ChatGPT, with a guide on how to use it well, to prepare Year 7 and 8 science lessons. The other did not use generative AI. According to the evaluation report, teachers in the ChatGPT group spent about 56.2 minutes a week on lesson and resource preparation, compared with 81.5 minutes in the other group, a saving of 25.3 minutes a week.
Did quality suffer?
An expert panel compared sampled lesson resources from both groups without knowing which had been made with ChatGPT, and found no evidence of a difference in quality. NFER's research director Helen Poet noted that the sample of resources was limited, so this result needs caution. The trial also found that teachers mostly used ChatGPT for one or two parts of a lesson, most commonly for creating questions or quizzes and finding activity ideas, rather than for the whole plan.
So the time saving was real, and quality held up in a group of teachers who had a guide and kept checking. That last part matters.
What do official guidelines say teachers must do?
The UK Department for Education's policy paper on generative AI in education is direct: "Any content produced requires critical judgement to check for appropriateness and accuracy." UNESCO's Guidance for generative AI in education and research calls for a human-centred approach and for protecting data privacy. Neither is Indian policy, but the principle carries over: the teacher stays responsible for what is taught.
A worked example: checking an AI-generated physics question
Suppose you ask for five free-fall questions for Class 9 and get this one back:
"A stone is dropped from a height of 20 m. Taking g = 10 m/s², it takes 4 s to reach the ground. Find its speed on impact."
Check it before it goes on a worksheet:
Recompute every number. For an object dropped from rest, h = ½gt², so t = √(2h/g) = √(2 × 20 / 10) = √4 = 2 s, not 4 s.
Check the follow-on answer. Speed on impact is v = gt = 10 × 2 = 20 m/s. You can confirm with v² = 2gh = 2 × 10 × 20 = 400, so v = 20 m/s. If the AI's answer key said 40 m/s, it used the wrong time.
Check the given data are consistent. A question that states both the height and the time gives students conflicting information. Remove the time, or ask for it.
AI chatbots can make arithmetic and reasoning errors in physics. Our post on what to do when AI gets physics wrong gives more examples.
A checklist for every AI lesson draft
Science. Is every statement correct? Watch for old models presented as current, and for common misconceptions, such as "heavier objects fall faster", slipping into explanations.
Numbers. Recompute every worked answer and check units.
Level. Does the vocabulary and maths match your class? A Class 8 worksheet should not need calculus.
Syllabus. Does it match what your board actually prescribes? Compare with the current NCERT textbooks or your board's syllabus for the year.
Sources. If the draft cites a study, a statistic or a quote, check that it exists. Our guide to AI and made-up sources explains why.
Privacy. Do not paste student names, marks or other personal data into a chatbot.
Purpose. Does each activity make students think, or only keep them busy?
Where can teachers cross-check a worked solution?
When you want a second line of reasoning to compare with an AI-generated answer key, the Fermi Problem Library has worked problems with a step-by-step solver that explains the reasoning rather than printing an answer. Use it as a comparison, and still recompute the numbers yourself.
Before your next class, take one AI-made worksheet and recompute every answer on it by hand.
