AI in Reading Instruction: Drawing Out Deep Reading Instead of Plot Summaries

Bring AI into a reading unit and one thing happens first: students stop reading the book and finish the job with "summarize this for me." They have the plot in hand, but their reading ability has not grown at all. Set AI up as a summarizer and it replaces reading; set it up as a questioner and it deepens reading. In reading instruction, AI's place is not reading in the student's stead but helping the student read better.

Have it generate questions instead of summaries

The key is to keep AI from telling students what is in the book. Give it the role of generating questions about the part the student has read instead.

• Prediction questions: after reading chapter one, the student asks for "two questions about what will happen next."
• Inference questions: "make me a question about the feeling behind this character's action that isn't stated directly" draws out reading between the lines.
• Connection questions: "make me a question that links this story to my own experience" connects the text to life.

Whether a student read the book shows up not in whether they know the plot but in whether they can pose a good question.

Students go one level deeper into the book by answering AI's questions in their own words. What matters is that the student writes the answer, not AI.

Running it across a semester

Take a 10th grade "one book a semester" reading program as an example, and you can run it like this.

1. Opening weeks: looking only at the cover and the table of contents, get "five questions you'd be curious about in this book" from AI to build motivation to read.
2. Middle weeks: each chapter, the student picks a key sentence themselves, then gets "a counter-question that examines why this sentence matters" from AI, stacking up material for discussion.
3. Closing weeks: with the student's draft response paper in hand, get check questions on "any perspective I left out."

In classrooms that adopted this approach, response papers not only got longer but the cookie-cutter, seemingly copied reflections dropped noticeably. Different questions produced different answers. Even reading the same book, each student was drawn to a different question, so the texture of the work was clearly unlike the days when everyone recited the same plot.

There is one thing to watch for as you run it. Among the questions AI produces, generic ones that can be answered without reading the book get mixed in. A question like "what kind of person is the protagonist?" is far too vague. If the teacher tells students to choose only questions tied to a specific scene or sentence in the book, you can prevent answering without reading. The judgment involved in choosing a question is itself part of reading ability.

Key takeaways

In reading instruction, AI should be a tool that deepens reading rather than one that does the reading. Block the summaries and have it generate prediction, inference, and connection questions, and students read actively as they answer. Don't have them ask for the plot; have them build good questions. Try this approach first in an activity where you read one book slowly together over a semester.