Because pretending AI doesn’t exist isn’t a plan.
Every teacher I’ve spoken with over the last year has asked the same question in one form or another:
“How do I assess students fairly when AI can do half the work for them?”
Underneath that question is something bigger than cheating. It’s about what a piece of student work even proves anymore.
My book makes an argument I come back to constantly: assessment had cracks in it long before AI showed up. AI just made the cracks visible. Once you see where they are, you can build something sturdier in their place.
A handful of teachers are tackling assessment differently.
Grade the thinking, not just the finished piece
A finished essay takes ten minutes to produce with a chatbot open in another tab. A student’s rough notes don’t. A voice memo where they talk through why they picked one argument over another doesn’t either. So collect that instead. Drafts. The prompts a student typed into the AI tool and what they did with the output. A short recording where they explain a place they got stuck. Edits made after feedback, with a note on what changed and why.
It sounds like more paperwork. Mostly it’s just pointing your attention somewhere new.
Build around something AI has never touched
A chatbot writes convincingly about categories. It knows nothing about your street, your family, or the argument you had with your neighbor last Tuesday. So send students outside the categories. Have them interview someone who lived through the event they’re studying. Ask them to tie a concept to something that happened to them last month. Send them to observe a specific place and report back on what they actually saw there.
AI can gesture at emotion in the abstract. It has never lived anywhere, and that gap is hard to close.
Put AI inside the assignment instead of policing its edges
Some of the teachers I’ve talked to stopped treating AI like something to confiscate. They hand it to students on purpose, as part of the task. Get three different answers from a chatbot, then explain which one is wrong and why. Ask it to critique a paragraph, then say what it caught and what it missed. Use it to generate a data set, then go analyze that data by hand.
Students learn to read machine output with a critical eye. That skill turns out to matter more than any ban ever did.
Ask for transfer, not recall
Memorize it, repeat it back. That was the whole game for a long time, and AI ended it in a single year. What’s left is harder to teach and more worth teaching. Can a student take something they learned in one setting and apply it somewhere unfamiliar? Can they defend a decision after you’ve changed the conditions on them mid-conversation?
Business schools have run whole programs on this kind of question for decades. It holds up because it resists automation in a way a fact sheet never could.
Talk to students, briefly, out loud
A two-minute conversation sometimes tells you more than a five-page paper. Ask a student to walk you through their reasoning. Ask what they’d change with another week. Ask where the AI helped and where it steered them wrong.
These conversations are hard to fake on the spot. Teachers keep telling me the same thing: within half a minute they can tell whether a student understands the material or just turned in something that reads well.
Draw the line before students have to guess at it
Students already know AI exists and already know how to use it. What they usually don’t know is where a particular teacher draws the line for a particular assignment. So tell them. What kind of help is fine. What needs to be cited. What has to be entirely their own from start to finish.
Most students aren’t hunting for a loophole. They just want the rule stated before they break it without meaning to.
The bottom line
None of this makes grading faster. Banning a tool and hoping the problem disappears would be faster. This is slower, and it takes more from the teacher up front. What it gives back is a clearer picture of what a student can actually do, which is the only thing assessment was ever supposed to measure in the first place.
For the full framework, with templates, implementation guides, and the science behind the process, check out The AI Teaching Revolution.