If a chatbot can complete an assignment in thirty seconds, then the assignment mostly measures whether a student pasted the question into a chatbot. That is not a reason to give up on take-home work. It is a reason to ask what the work is for, and to design it so that the thinking is still visible.
Research on how students actually use AI is sobering. A 2026 preprint, "Your Students Don't Use LLMs Like You Wish They Did", analysed real student conversations with AI tutors and found that students mainly use them to extract answers, and that when the tool was built into a course they often pasted assignment questions in directly. A USC report found most college students use tools like ChatGPT to get quick answers rather than to learn concepts, unless professors guide them to deeper use. In an NPR/Ipsos poll reported by K-12 Dive, 55% of teachers said students use AI mostly as a shortcut to avoid schoolwork, and 54% said it makes critical thinking harder to develop. Whatever your own view, your students are likely to have it in their pockets.
Start with the purpose of the task
Before changing anything, ask: what do I want students to be able to do after this?
- If the goal is knowing facts or methods, the evidence needs to come from something students do without a tool, such as a short in-class check.
- If the goal is building an argument or a product, the evidence can be the process: notes, drafts and decisions.
- If the goal is learning to use AI well, the task should include the AI conversation and a reflection on it.
Different goals need different designs. One rule does not fit every assignment.
Seven design moves that help
1. Make the process visible
Ask for a plan, a draft and a revision, each with a short note explaining what changed and why. A single finished product hides everything you want to see.
2. Add a short in-class component
A five-minute explanation, a few unaided questions or a brief conversation shows whether the take-home work is the student's own understanding. It does not need to be graded heavily.
3. Use your own class as the source
Tasks that draw on a specific class discussion, a local example or a text you studied together are harder to answer generically. They also tend to be more interesting.
4. Ask for reasoning, not just answers
"Solve this and explain why each step is valid" is more revealing than "Solve this." Ask students to say what they tried that did not work.
5. Build in personal or local context
Ask students to connect a concept to their own experience, community or interests. This is better for engagement as well as for authenticity, so do not treat it only as an anti-AI trick.
6. Shorten and scaffold
A large one-off assignment invites last-minute shortcuts. Break it into stages with feedback in between, so that students are learning along the way.
7. Use oral checks
A two-minute conversation ("Talk me through how you got this") tells you a great deal, and is difficult to fake.
Set clear rules for AI use
Students are more likely to follow rules that are specific. Consider a simple three-level system and label each assignment:
| Level | What it means | Example |
|---|---|---|
| No AI | The work must be done without AI tools | In-class writing, quizzes |
| AI for support | Use for ideas, feedback or explanation, but not for producing submitted text | Brainstorming, checking a draft |
| AI allowed, with a record | Use it as you like, and submit the conversation and a reflection | Research projects |
A common complaint is that students receive different rules from different teachers, and get confused. An EdWeek Research Center survey found that 73% of educators say their district has no districtwide AI initiative, so classroom experiences vary widely. Agree a common framework with colleagues if you can, and put the level at the top of every task.
Handle suspicion with care
Do not rely on a detection tool alone to accuse a student. Such tools can be wrong, and a false accusation is costly. If you are concerned, start with a conversation: ask the student to explain their work, walk through how they wrote it, or complete a short related task in front of you. Compare with their in-class work over time. Follow your school's process.
A worked example: a history essay
Old task: "Write an essay on the causes of the French Revolution."
Redesigned:
- In class (10 min): students list three causes from memory, unaided.
- Take-home plan: an outline, with a note on which cause is the strongest and why.
- Take-home draft: with a short reflection: "What did I change after feedback?"
- In class (5 min): each student explains their thesis aloud to a partner.
- Final: submitted with an AI-use note, if AI was permitted.
The essay is still there, but it now sits inside a process that shows the student's thinking.
Talk with students about why
Explain your reasoning. Students who understand that practising writing or problem-solving is how they get better at it are more likely to take that practice seriously. Present it as a skill they will need after school, not as a rule to obey.
Summary
- Ask what each task is for, and design to fit.
- Make the process visible and add a short unaided component.
- Label every assignment with a clear AI-use level.
- Talk with students before accusing them.
Related guides
- How to write a school AI policy
- How to write a rubric
- Teaching students to ask better questions
- Education News: latest research, a live feed of new education studies