Hide a marker in the assignment.
Let the chatbot walk into it.
You add a few lines of text to your assignment that a student reading the page can never see — but a chatbot fed the whole page reads and obeys them. If those instructions show up in an essay, the page went through a chatbot.
Put in the assignment exactly as your students will get it. Nothing leaves your browser.
You get the same sheet back with hidden text added. Copy it into Google Docs, Word or your course page and hand it out as usual.
If an essay follows instructions no human could see, it shows up here as a triggered check.
Useful reading
Short articles on how this technique works in real classrooms and what to watch out for. Links open in a new tab.
- ↗
From Trojan horses to AI-proof exams: how professors are tackling students’ AI use
How faculty worldwide use hidden-prompt techniques like these — and their limits.
Nature
- ↗
Professor creates trap in his midterm to catch students using AI
The real case this technique became famous for — 32 submissions tripped one hidden line.
NBC / TODAY
- ↗
Generative Artificial Intelligence — teaching guidance
Syllabus language, assignment design and policy advice to pair with these checks.
Cornell Center for Teaching Innovation
- ↗
Prompt injection, explained
A readable series on why hidden instructions work on chatbots — and why they are not foolproof.
Simon Willison
Step 1 · Paste your assignment
Copy the assignment text and paste it in the box below. 0 words so far.
Step 2 · Pick your checks
Click a card to switch it on or off. The four suggested ones are already on — if you are not sure, leave them as they are.
Step 3 · Build the sheet
Everything here runs in your browser · no AI, no credits · nothing is stored. Check your institution's academic-integrity policy before using hidden text in graded work, and treat a triggered check as evidence to discuss with the student, not an automatic verdict.