A page can feel familiar while the idea remains difficult to explain. The teach-back tutor starts with your explanation, then uses a focused question to find the place where your understanding becomes uncertain.
Give the AI a trusted source and make your own attempt before asking for help. The point is to practise retrieving and connecting ideas, rather than read another fluent explanation immediately.
The prompt
Be my teach-back tutor for [TOPIC]. Use these learning materials as the reference: [PASTE OR ATTACH THE RELEVANT MATERIAL] First, ask me to explain the topic in my own words without looking at my notes. Wait for my explanation; do not give a model answer first. Compare my explanation with the supplied material. Briefly identify something I explained accurately, then ask one question about the most important gap, ambiguity, or misconception. Do not reveal the answer yet. After I respond, give concise feedback tied to a specific passage or section of the material. If I am stuck, offer one hint before the full explanation. Do not invent references. Flag anything the material does not establish. Repeat for no more than three questions. Finish with two points to review and one new application question for a later practice session. Do not include its answer unless I ask.
Make the first attempt yours
Choose a small enough topic to explain in a minute or two: why seasons change, how a particular spreadsheet formula works, or the argument in one section of an essay. Uploading an entire textbook and asking to be tested on everything makes it harder to focus the session.
Close your notes while you explain. Include what you are unsure about instead of quietly looking it up. A rough attempt gives the tutor something useful to respond to. Copying the source into your answer mostly tests whether the model can recognize its own input.
For example, a learner might describe an argument accurately but leave out the evidence connecting two claims. A helpful follow-up would ask them to explain that connection. It should not replace their entire answer with a polished summary and call the exercise finished.
Use feedback, then return later
The approach draws on retrieval practice. MIT Open Learning describes how recalling knowledge, particularly with feedback, supports retention. That research supports the learning principle; it does not establish that this particular AI prompt has been tested or that every correction it gives will be accurate.
Check disputed feedback against the supplied material. A missing detail may be important, optional, or simply outside the question you were answering. Ask the tutor to point to the relevant passage and explain why the omission matters. If the source itself is unclear, keep that uncertainty visible.
Save the final application question for another day. Trying it later, before reopening your notes, gives you a different challenge from repeating an answer you just read. For a quick session, reduce the limit to one follow-up question. End with something specific to revisit, rather than a vague score for how well you understand the subject.