One AI Teaching Practice to Keep, One to Change, and One to Test 

teacher stressed with AI lesson planning
  • Home
  • /
  • Insights
  • /
  • One AI Teaching Practice to Keep, One to Change, and One to Test
October 16, 2026

For a lot of teachers, AI in the classroom looks like a pile of small, unresolved decisions made alone, late at night, with no clear rules to lean on.

Fortunately for you, you spent a full semester running experiments, whether you meant to or not. Some worked. Some made things worse. One or two deserve an intentional trial.

The trick is to stop treating AI as one giant question and start treating it as a handful of specific practices. Keep the ones that earn their place. Fix the one that keeps hurting you and your students. Test one new idea, with a clear way to judge whether it worked.

Keep Asking AI to Tear Apart Your Lesson Plan Before Your Students Do

Most talk about AI for teachers focuses on writing lesson plans faster. The smarter move is to use it to attack them.

Finish your plan the way you always do, with your objectives, your activities, and your timing. Then paste it into AI and ask it to respond as three different people: a confused student, a bored student, and a skeptical department chair.

The confused student finds where the directions fall apart. Meanwhile, the bored student spots the 10 minutes where your energy sags and the phones come out. And the skeptical chair asks the question you've been avoiding about how this lesson actually connects to the standard.

Every flaw they catch on Sunday night is a flaw your students never have to sit through on Monday morning.

This practice protects your judgment. You still write the lesson, AI simply plays the critic you never have time to find.

Build Your Three Critics Once

Write the three role prompts one time and save them. Spell out who each person is, like this: "You are a ninth grader who missed last class and is lost. Read this lesson plan and tell me exactly where you would get stuck." A saved prompt takes 30 seconds to reuse. One you rewrite every week will get skipped by November.

Then run every new lesson past all three. Skipping the department chair is tempting, but that's usually the critic with the most useful complaint.

Let Your Judgment Have the Final Word

AI will invent problems. Some critiques will be generic or flat wrong. Use your expertise to decide what changes, and don’t just put into the plan just because a chatbot said so.

Keep student information out of the process entirely. Use the plan itself, never names, grades, or anything that could identify a real student.

Learn Which Critic Earns Its Keep

After each lesson, jot one line noting what the AI critics flagged and whether your actual students hit the same wall. Within a few weeks, you'll know which critic deserves your attention and which prompt needs sharper wording.

Share the prompts with one colleague, too. A practice that only you use can vanish during your first busy week. With two teachers using it, you have someone to compare notes with when a critique misses the mark.

Stop Trusting the AI Detector and Start Asking Students to Talk You Through Their Work

AI detectors guess. They flag honest writers, especially students who write in a plain, careful style or who are still learning English. They also miss generated work once a student runs it through a rewording tool. You end up holding a score you can't explain to a student, a parent, or an administrator.

Worse, the detector turns you into a prosecutor. Every essay becomes a case to build. Trust between you and your students wears away one flagged paper at a time.

Here's the practice to change. Drop detector scores as proof of anything, and replace them with a short conversation where each student walks you through one paragraph they wrote.

Ask the Same Three Questions Every Time

Pick a paragraph at random and ask the same three questions: why they chose this point here, what they cut to get to this version, and what their source actually said.

A student who wrote the paragraph can answer all three without breaking a sweat. A student who pasted it usually stalls on the second question, because there was never an earlier version to remember.

Keeping the questions identical matters. It makes the conversation fair, and it keeps your own suspicions from steering where it goes.

Talk With Every Student So No One Feels Singled Out

Hold these conversations with the whole class, at least once per major assignment. When everyone does it, it feels like part of the course instead of an accusation. Students also start writing differently once they know they'll have to explain their choices out loud. That shift alone is worth the time.

Write One Line After Each Conversation

Jot a single note about what the student explained well and where they struggled. Over a term, those notes paint a far clearer picture of each writer than a stack of percentages ever could. If a real integrity concern does come up, you'll have specific, fair evidence drawn from the student's own words.

Test Handing Students an AI Essay and Your Rubric Before They Write a Word

Students see your rubric on day one. Most of them never really read it.

That changes the moment they have to use it on someone else's work. So, here's the experiment: before students draft their own response to an assignment, hand them an AI-written version of it along with your actual rubric. Their job is to grade it, score every row, and name exactly what it got wrong.

Something shifts when students sit in the grader's chair. The AI essay sounds confident. It's tidy and grammatical. Then they start checking it against the rubric and find the thin evidence, the vague claims, and the conclusion that says nothing at all. Suddenly "polished" and "good" stop meaning the same thing.

Set Up the Test Before You Run It

  • Generate the essay yourself, then plant three or four flaws tied directly to rubric rows, like a misquoted source or a claim with no support.

  • Pilot it in one unit, with one assignment, so you can see the effect clearly.

  • Decide in advance what a strong critique looks like, and write down the flaws you expect students to catch.

  • Use a synthetic prompt and essay, never a real student's work.

Measure It Against Last Term

This is a test, so treat it like one. Compare final drafts from this unit with the same assignment from last term. Look closely at the rubric rows where the AI essay was weakest. If students improved there, you have something worth keeping.

Also pay attention to the critiques themselves. Students who catch only grammar issues need more practice reading for substance. Those who catch the planted flaws, plus a few you missed, are telling you the exercise works.

Give it one unit. Then decide, with real evidence, whether it earns a permanent spot in your course.

Spend Five Minutes Every Friday Writing the AI Log You'll Thank Yourself for in May

Picking one practice to keep, one to change, and one to test takes effort right now. Jot them down, with each entry answering the same five prompts:

  1. What AI did this week, in one sentence.

  2. Roughly how much time it saved or cost you.

  3. What broke or fell flat.

  4. How students reacted.

  5. One thing to try next week.

Turn the Log Into Next Term's Decisions

Set a recurring Friday reminder so the habit doesn't depend on willpower. At midterm, read through the log with one colleague. Patterns jump out fast when someone else is reading alongside you.

By May, the hard part is already done. Your next practices to keep, change, and test will be sitting right there in your own words, backed by a full term of evidence instead of a late night guess.

Bizzuka builds AI training and workshops around the way your educators actually teach. Sessions are hands-on: faculty write their own critic prompts, practice the student conversation, and draft clear AI expectations for real assignments from their own courses. Everything fits your institution's policies, subjects, and comfort levels with AI. Faculty walk out with practices they can use in class the following Monday.

Your next term doesn't have to start with another late night guess. Let's talk about what your educators need. Schedule a conversation with us to explore AI training options and workshops built for your institution.