AI Exit Tickets That Actually Inform Tomorrow's Lesson

Exit tickets are one of the most efficient formative assessments a teacher has — three minutes at the end of class, one honest signal. The problem has never been collecting them. It's reading 140 of them before you plan tomorrow.

This guide shows a five-minute AI workflow that turns raw exit tickets into a plain-English brief: what students got, what confused them, which misconceptions surfaced, and one concrete adjustment for the next lesson.

The 3-question exit ticket

Keep the prompt small and specific. Three questions is the sweet spot — enough signal, not enough to skim.

1. One thing you can now do that you couldn't at the start of class.
2. One thing you're still fuzzy on.
3. On a scale of 1–5, how confident are you doing tonight's homework alone?

Collect on paper, a Google Form, or a shared doc — whatever your students already use. If it's a form, you already have a CSV; if it's paper, snap a photo of the stack and paste text into the AI. Either way, the input is a list.

The AI prompt

Paste the responses into any general-purpose AI (Gemini, ChatGPT, Claude) with this framing:

You are helping a [grade + subject] teacher plan tomorrow. Below are 3-question exit tickets from today's lesson on [topic]. Do four things: (1) list the top 3 things students report they can now do, (2) list the top 3 sticking points in students' own words, (3) name any misconception that appears in more than 2 responses, (4) suggest one 10-minute reteach activity and one extension for students who reported 5/5 confidence. Keep the whole response under 300 words.

The important design choices in that prompt:

• Role + audience ("helping a 7th-grade science teacher plan tomorrow") anchors the register.
• Numbered outputs stop the AI from writing an essay.
• "Students' own words" forces it to quote instead of paraphrase into blandness.
• A word cap keeps the brief scannable during passing period.

What a good output looks like

For a lesson on the water cycle, a real return might read:

Got it: naming the four stages, drawing arrows in the right direction, connecting evaporation to sun energy.

Sticking: "condensation and precipitation feel the same," "why does water go up in the first place," "I can't tell where the energy comes from."

Misconception (7 students): clouds are made of water vapor, not liquid droplets.

Reteach (10 min): ice-cold glass demo — students touch, name the phase change, then re-draw one arrow in yesterday's diagram.

Extension: predict how the cycle would change on a planet with no sun. Written response, one paragraph.

That's enough to redraw tomorrow's opener in about 60 seconds. You do not need a polished lesson plan; you need to know what to open with.

Guardrails

Exit tickets are student data. A few rules that keep the workflow safe and honest:

• Strip names before pasting. Most spreadsheet tools do this in one click ("hide column"). Free-tier AI products should not see student identifiers.
• Don't let the AI grade. It's summarizing signal, not making judgments about individual students. Averaging confidence scores is fine; naming a student is not.
• Read the summary before you trust it. If a "sticking point" doesn't match what you saw in class, the AI probably over-generalized from one loud response. Skim the raw tickets for that item.
• Save the raw tickets. The summary is disposable; the tickets themselves are the record of the lesson.

Why this beats reading them one by one

A teacher's time on exit-ticket triage should be spent making a decision, not decoding handwriting. AI is good at pattern-matching across short responses and terrible at the professional judgment that follows. The workflow above uses each side for what it's good for: the AI finds the pattern, you decide what to do with it.

Run it once. If tomorrow's reteach feels sharper than usual, keep the prompt. If not, tweak the four numbered asks — the framing is the whole game.