Here's a test for you: Take one of your recent assignments. Copy the instructions into ChatGPT. See what happens.
Did it produce an A-level response in 30 seconds?
Congratulations. Your assignment just became obsolete.
The goalposts have moved. Tasks that once required deep thinking can now be completed by AI without any thinking at all. But we have a 70-year-old framework that tells us exactly how to design AI-resistant assignments: Bloom's Taxonomy.
Bloom's Taxonomy: A Quick Refresher
Benjamin Bloom and colleagues developed this framework in 1956, revised in 2001. It categorizes cognitive skills into six levels, from lowest to highest:
- Remember — Recall facts and basic concepts
- Understand — Explain ideas or concepts
- Apply — Use information in new situations
- Analyze — Draw connections among ideas
- Evaluate — Justify a decision or course of action
- Create — Produce new or original work
For decades, educators used this taxonomy to design progressively challenging assignments. Then AI arrived. And it scrambled everything.
What AI Can Do at Each Level
According to research published in Taylor & Francis (2024), AI's capability varies dramatically across Bloom's levels. Understanding these capabilities is essential for designing effective assessments.
Level 1: Remember (AI Dominates - 95%+) — This level requires retrieving relevant knowledge from memory. AI can instantly recall any fact in its training data, list definitions, dates, formulas, vocabulary, and generate comprehensive timelines and fact sheets. Reality check: If your assessment only requires remembering, AI will ace it.
Level 2: Understand (AI Excels - 85-90%) — This level requires constructing meaning from information. AI can summarize complex texts with remarkable accuracy, explain concepts in multiple ways, translate between languages and reading levels, and generate examples and analogies. Reality check: AI can explain most concepts better than many students.
Level 3: Apply (AI Is Competent - 70-80%) — This level requires using knowledge in a new situation. AI can solve standard problem types (math, physics, coding), apply formulas and algorithms, and execute procedures step-by-step. Reality check: For well-defined problems with clear procedures, AI performs well. But here's where we start seeing cracks.
Level 4: Analyze (AI Shows Limitations - 50-60%) — This level requires breaking material into parts and determining relationships. AI struggles with deep contextual understanding of nuance, recognizing unstated assumptions, detecting subtle biases, understanding historical or cultural context, and connecting across widely disparate domains. Reality check: This is where strategic assignment design starts winning.
Level 5: Evaluate (AI Has Significant Gaps - 35-45%) — This level requires making judgments based on criteria and standards. AI cannot make nuanced ethical judgments, evaluate based on lived experience, weigh competing values authentically, or defend judgments under questioning. Reality check: AI can suggest evaluations, but can't authentically judge.
Level 6: Create (AI Is Weakest - 25-35%) — This level requires producing original work with a unique perspective. AI cannot create from genuine human experience, produce work with authentic voice, generate truly paradigm-shifting ideas, or defend creative choices under deep interrogation. Reality check: AI can mimic creativity, but it can't truly create.
Key insight: AI capability decreases as you move up Bloom's Taxonomy. Your assignment design should exploit this.
The New Bloom's: Designing AI-Resistant Assignments
Levels 1-2: Remember & Understand
The Problem: AI excels at recall and explanation.
The Solution: Require students to connect information to their personal experience or specific context.
AI-Vulnerable: "Define photosynthesis."
AI-Resistant: "Observe a plant in your home for one week. Document its changes with photos and explain the photosynthesis process you're witnessing."
Why it works: AI doesn't have personal experiences or access to students' specific environments.
Level 3: Apply
The Problem: AI can apply procedures to standard problems.
The Solution: Create application tasks that require working with novel, specific, local contexts.
AI-Vulnerable: "Calculate the area of this rectangle."
AI-Resistant: "You're redesigning our classroom to create a collaboration zone. Measure the space available, calculate how many 4-person tables will fit with 3 feet of spacing, and justify your layout choice."
Why it works: AI doesn't have access to your specific classroom dimensions.
Level 4: Analyze
The Problem: AI can perform surface-level analysis.
The Solution: Require students to analyze from multiple stakeholder perspectives and explain their analytical process.
AI-Vulnerable: "Analyze the causes of World War I."
AI-Resistant: "Analyze the causes of WWI from three perspectives: a German factory worker, a British suffragette, and an Ottoman diplomat. Which perspective reveals causes that textbooks typically miss? Explain how your own background influenced which perspective you found most compelling."
Why it works: Requires metacognition and explicit consideration of how perspective shapes analysis.
Level 5: Evaluate
The Problem: AI can apply evaluation criteria mechanically.
The Solution: Create evaluation tasks with competing values, ethical dilemmas, or ambiguous situations.
AI-Vulnerable: "Evaluate which energy source is best: solar, wind, or nuclear."
AI-Resistant: "Your town has budget for ONE renewable energy project. Solar would benefit 500 homes immediately. Wind would benefit 2,000 homes in 5 years. Evaluate which project deserves funding, addressing: environmental impact, economic justice, long-term sustainability, and community preferences. Defend your recommendation knowing half the town will disagree."
Why it works: No objectively "correct" answer; requires weighing competing values and defending judgment.
Level 6: Create
The Problem: AI can generate creative content.
The Solution: Require work grounded in personal experience with visible iteration and authentic voice.
AI-Vulnerable: "Write a persuasive essay on climate change."
AI-Resistant: "Create a persuasive campaign to change one specific behavior in our school community. Include: audience research (survey at least 10 students), 3 creative artifacts (poster, video script, social media plan), and a reflection explaining how you adapted your approach based on peer feedback."
Why it works: Requires lived experience, iteration based on real feedback, and authentic reflection.
The "AI-Proof Assignment" Checklist
Use this checklist to evaluate any assignment:
- ☐ Personal Context Required — References specific experiences or environment
- ☐ Novel Application — Unique enough that AI hasn't seen this exact scenario
- ☐ Multiple Perspectives — Must consider and synthesize multiple viewpoints
- ☐ Metacognitive Reflection — Explain thinking process, not just final answer
- ☐ Ambiguity Present — Genuine uncertainty or competing values
- ☐ Iteration Required — Must revise based on feedback
- ☐ Authentic Voice — Detectable as this specific student's perspective
- ☐ Oral Defense Possible — Could explain and defend in conversation
- ☐ Process Visible — Thinking journey shown, not just product
- ☐ Physical Evidence — Requires being present (measurements, photos)
Scoring: 8-10 checks = AI-Resistant. 5-7 checks = Moderately Resistant (add elements). 0-4 checks = AI Completes Easily (redesign needed).
Case Studies: Assignments Redesigned
Case Study 1: History Essay → Multi-Perspective Analysis
Original (AI-vulnerable): "Write a 5-paragraph essay explaining the causes of the American Revolution." AI Test: ChatGPT produced an A-level essay in 45 seconds.
Redesigned (AI-resistant): "You're a museum curator designing an exhibit on the American Revolution. Create exhibit labels (200 words each) for three artifacts from different perspectives: a wealthy Boston merchant, an enslaved person in Virginia, and a Mohawk leader. For each, explain what this artifact reveals about revolution causes and what this perspective shows that textbooks miss."
Case Study 2: Science Lab Report → Process Documentation
Original (AI-vulnerable): "Complete the photosynthesis lab. Write a report with hypothesis, procedure, data, and conclusion." AI Test: ChatGPT generated a perfect lab report without doing the lab.
Redesigned (AI-resistant): "Conduct the photosynthesis lab. Create a 'scientific vlog' (3-5 minutes) showing: Your initial hypothesis and reasoning (before the lab), unexpected results and how you responded, one thing that went wrong and how you troubleshot, what you learned about scientific thinking."
Case Study 3: Math Problem Set → Real-World Application
Original (AI-vulnerable): "Solve these 20 quadratic equations." AI Test: ChatGPT solved all 20 instantly with perfect accuracy.
Redesigned (AI-resistant): "You're hired by our school's athletics department to optimize the shot put landing area. Using quadratic equations, determine optimal throwing angle and landing zone dimensions. Create a proposal with mathematical analysis, scale diagram, and explanation a non-math-expert would understand."
Teaching Students to Use AI Wisely
Don't ban AI from these assignments. Teach students how to use it strategically.
Instead of: "No AI allowed"
Try: "You may use AI as follows..."
Example AI usage policy:
You may use AI to:
- Find primary sources from different perspectives
- Summarize long historical documents
- Generate initial research questions
You may NOT use AI to:
- Write your exhibit labels (these must be your words)
- Make analytical claims about what perspectives reveal
- Write your reflection on how your thinking changed
Required: Attach an appendix showing what AI tools you used, what prompts you used, and how you verified AI information.
Why this works: It teaches AI literacy (students learn strategic vs. lazy AI use), maintains learning (the cognitive work stays with students), builds metacognition (students reflect on their process), and prepares for the future (this is how professionals use AI).
Your Action Plan
Step 1: Choose an Assignment. Pick something you'll use in the next 2-4 weeks.
Step 2: Run the AI Test. Copy instructions into ChatGPT. What happens?
Step 3: Identify Cognitive Level. Where does it fall on Bloom's? Lower or higher?
Step 4: Add 3 AI-Resistant Elements. Consider personal context, multiple perspectives, metacognition, ambiguity, iteration, oral defense, physical evidence, or process documentation.
Step 5: Rewrite and Test Again. Can ChatGPT still complete it easily? If yes, add more elements.
The Philosophical Question
Here's where we need to get uncomfortable.
For decades, we've asked students to do things AI can now do instantly: summarize texts, solve problem sets, write essays on standard topics.
Were we ever teaching them to THINK, or just to PERFORM?
If AI can "complete" your assignment, maybe your assignment was never testing real thinking in the first place. AI forces us to design assignments that require what AI cannot do—authentic human thinking. This means thinking that is grounded in personal experience, weighing competing values ethically, adapting to ambiguous, messy situations, reflecting on its own process, and creating something genuinely new.
This is harder to design. It's harder to grade. But it's the only teaching that matters now.
The Bottom Line
Bloom's Taxonomy isn't obsolete in the AI age. It's essential—but we must use it differently.
Old Approach: Move students up Bloom's pyramid from lower to higher-order thinking.
New Approach: Design every assessment at the higher levels of Bloom's that AI struggles with—and explicitly teach students to use AI for lower-level tasks.
The goal shifts from: "Can students complete all levels of Bloom's?"
To: "Can students use AI for lower-level work while demonstrating higher-order thinking that's uniquely human?"
When your assignments require authentic analysis, ethical evaluation, and creative synthesis grounded in human experience, you're teaching the only skills that matter in an AI world.
References
- Taylor & Francis (2024). "Promoting cognitive skills in AI-supported learning Environments: the integration of Bloom's Taxonomy."
- Anthology (2025). "Reframing Bloom's for the Age of AI."
- Online Learning Consortium (2025). "Using Bloom's Taxonomy to understand AI Adoption in Education."
- SAGE Journals (2025). "Can AI Generate Questions Aligned with Bloom's Revised Taxonomy?"
- Stanford Teaching Commons (2024). "Increasing Student Engagement."
- Anderson, L. W., & Krathwohl, D. R. (2001). A taxonomy for learning, teaching, and assessing.
- Bloom, B. S. (1956). Taxonomy of educational objectives.
Continue Reading
- The Cognitive Offloading Crisis: Why 75% of Students Are Using AI Wrong — Understand the neuroscience behind why students are using AI in ways that harm learning.
- Beyond Plagiarism Detection: Rethinking Academic Integrity in the AI Era — Move from detection to design-based approaches for preserving academic integrity.