Three kinds of tool worth telling apart

Adaptive practice tools choose the next question based on how a pupil answered the last one. They have existed for years and usually work within a fixed bank of questions written by people, so the content is predictable.

Generative tools, such as chatbots built on large language models, produce new text, images or explanations on request. They are flexible, but they can state wrong information with confidence, and their answers change each time.

Feedback tools sit in between. They comment on a pupil's essay, code or pronunciation. Their value depends on whether the feedback is specific enough for a child to act on without a teacher translating it.

The same app can contain more than one of these. A reading platform might use adaptive levelling for book choice and a generative assistant for comprehension questions. Look at each feature separately, because the risks differ: an adaptive quiz rarely produces wrong content, while a generative assistant might. Parents can apply the same test at home, asking which part of an app is fixed content and which part writes new answers.

Questions to ask before adopting a tool

A short checklist saves time later and gives school leaders something consistent to apply across departments.

• What data does it collect about pupils, where is it stored, and can it be deleted?

• Does it meet the age requirements in its own terms of use for the year group you teach?

• Can a teacher see what pupils asked and what the tool replied?

• Does it support a learning goal you already have, or does it add a new activity with no clear purpose?

• What happens to the lesson if the tool is unavailable?

A first trial in class

Try a tool with a small, low-stakes task before building it into a unit. With a Year 8 class (ages 12 to 13), spend 30 minutes on a science topic such as the water cycle. Materials: one shared device per pair, a teacher-approved chatbot or explainer tool, and a printed sheet with three columns headed "The tool said", "We checked in the textbook" and "Correct?".

Pairs ask the tool to explain one stage of the cycle, record its answer, then check it against the class textbook. The last ten minutes go on a class discussion of any errors found. Pupils learn the content and, just as usefully, learn that an AI answer is a starting point to be verified.

Key points

• Identify whether a tool is adaptive, generative or feedback-based before judging it.

• Check data handling and age limits before pupils log in.

• Start with a short, low-stakes trial that includes checking the tool's answers.

• Keep a plan for the lesson that does not depend on the tool.

Photos: Vitaly Gariev, Christin Hume / Unsplash