All activities Professional learning
Teachers run a glow-grow-question peer review on two AI-assisted drafts, questioning the AI's suggestions as hard as the writer's, then adapt the protocol for their own students.
When students revise with a generative AI tool, the tool's suggestions can quietly replace the most valuable feedback a writer gets: a real reader's question. This session puts peers back at the center. Participants act as students and give each other glow, grow, and question feedback on two fictional AI-assisted drafts. They also judge each AI suggestion (accept, adapt, or reject) and have to explain why. Then they debrief as teachers and adapt the protocol, with supports, for their own grade level. They finish by naming what a coach or leader would see in a classroom where this is working.
Hook
Ask: "Think of a piece of feedback that actually changed your work. Who gave it, and what made it useful?" Pairs share for one minute each. Collect a few answers and listen for the pattern: specific, from someone who read closely, often a question that made you rethink. Then ask: "Could a chatbot have given you that? Why or why not?"
Facilitator noteMost rooms land on this: useful feedback comes from a reader who cared what you meant. Hold on to that; it's the reason peers stay in the loop even when AI is available.
Model
With a volunteer, model the protocol on Draft 1 (Handout A) while others watch. The writer reads the draft aloud. The partner gives one glow, one grow, and one question, using the stems. Then you both look at the three AI suggestions and decide together: accept, adapt, or reject, and why. Narrate your thinking on Suggestion 3: "The AI added a statistic the writer never had. Where did it come from? I can't trace it, so it can't go in."
Facilitator noteDraft 1, Suggestion 1 (add an example) is worth adapting: the idea is good, but the student should add their own example. Suggestion 2 (swap in fancier vocabulary) makes the writing less clear for a fifth-grade reader; reject or adapt. Suggestion 3 invents an unsupported claim; reject.
Practice
Pairs work on Draft 2 as if they were high school or college students. Partner A is the writer and reads aloud; Partner B gives glow, grow, and question. Then they switch roles for the AI's suggestions: B reads each suggestion aloud and A, as the writer, decides whether to accept, adapt, or reject it while B pushes back with a question stem. Both log their decisions on Handout B.
Facilitator noteLook for pairs who ask the AI questions they'd ask a human reader: "What makes this counterargument stronger?" "Would my actual audience say this?" Suggestion 2 on Draft 2 removes the writer's personal experience to sound "more academic". Watch whether pairs notice that it deletes the draft's best evidence.
Debrief
Pairs join into groups of four. Each group answers two questions on sticky notes, one color each: "What did the human partner notice that the AI suggestions missed?" and "What did the AI suggest that was actually helpful?" Post and cluster the notes. Then ask the room: "What did the writer have to know to judge the AI's suggestions? Where do students learn that?"
Facilitator noteThe honest answer usually includes some helpful AI suggestions. The point is not that AI feedback is bad; it's that judging it is a thinking skill, and a peer's questions help students build it.
Apply
Using Handout C, each person adapts the protocol for one upcoming assignment. They decide whether AI belongs in this task at all, and if it does, how it's used and disclosed (for students under 13, the teacher projects the tool and the class judges its suggestions together). They add at least two supports so every student can give and receive feedback: sentence stems, a sketch or audio option, feedback in a home language, partners chosen with care, a shorter draft. Then they name what students will turn in as evidence: the log and a short revision note saying what they changed and why.
Facilitator notePush on the evidence line. "Students will do peer review" is an activity; "Students turn in their log and a note naming one suggestion they rejected and why" is evidence of thinking.
Transfer
Partners swap planners and complete the last section together: "If a coach or principal walked in during this peer review, what would they see and hear that shows it's working?" Push past "students are on task" to things like "students quoting each other's drafts," "a student explaining why she rejected an AI suggestion," "revision notes that name a peer's question." Coaches and leaders in the room turn these into two or three look-fors they could use in a walkthrough.
Facilitator noteThis step connects classroom practice to how it's supported and evaluated. When teachers write the look-fors themselves, walkthrough feedback lands as support, not surveillance.
Run everything on paper: the drafts, the AI suggestions (already printed on Handout A), the log, and the adapter. With students, you can do the same thing. Print a draft with AI suggestions in the margin, or, for students under 13, project the tool, write its suggestions on the board, and have partners judge them on paper. The protocol itself needs no device.
Put the drafts in a shared document and have partners leave glow, grow, and question as comments, then reply to each AI suggestion comment with accept, adapt, or reject and a reason. The version history becomes evidence of revision. In an LMS, post the log as an assignment and collect revision notes with the final draft. Students 13 and older may use a district-approved AI tool to generate suggestions on their own drafts; they should never paste in a peer's work without permission.
Does it need a screen? A shared document with comments and version history shows who said what and what changed after each piece of feedback, which paper can't track as easily. The AI earns its place only when students have to judge its suggestions; the log is the evidence of that judgment.
What you should be able to see or collect if it worked.
Teachers practice peer feedback that judges AI revision suggestions with reasons, the same revising and questioning students do on their drafts.
Run your adapted protocol on the next writing or project draft. Collect the logs and revision notes, and bring three samples (names removed) to your PLC or coach to look for how many students rejected or adapted a suggestion with a clear reason.
