Create Differentiated Exercises
On this page
Differentiation at Scale
Differentiating exercises for thirty students with diverse profiles is a daily challenge. The difficulty isn’t conceptual; it’s logistical. Everyone knows they should create three versions of the same assignment, but no one has time to write them every week. ChatGPT removes this constraint by quickly generating variants of a single exercise tailored to different levels, working modes, and specific needs. However, you must decide exactly what varies. Without that prior decision, you’ll get three different exercises instead of three versions of the same one, making it impossible to grade them together or compare learning outcomes.
The Three Axes of Differentiation
Differentiation by content varies the complexity of the subject matter. At the foundational level, you work on basic concepts using simplified vocabulary and concrete, everyday examples. At the standard level, you follow the official curriculum with expected disciplinary vocabulary and varied examples. At the advanced level, you explore interdisciplinary connections, critical analysis, and open-ended problems. For proportionality, the range spans from scaling a cooking recipe to critiquing a news graph where the y-axis doesn’t start at zero.
Differentiation by process varies how the student works while keeping the subject identical. The guided mode provides detailed steps, progressive hints, and templates to complete. The semi-autonomous mode gives clear instructions with a few leads and allows a choice between several approaches. The autonomous mode poses an open problem, grants methodological freedom, and requires self-assessment. This is the most efficient axis for the classroom because it allows everyone to tackle the same question: collective correction remains possible, and no student feels relegated to a “watered-down” topic.
Finally, differentiation by product varies what the student must submit: structured writing (short answers, argumentative text, essay), visual support like a mind map, infographic, or annotated diagram, or oral production in the form of a presentation, debate, or podcast. This axis is often most effective for students who disengage from writing despite understanding the concepts, and it sometimes reveals that a notion thought to be unmastered was actually learned long ago.
Generate Three Versions at Once
The approach that saves the most time is requesting all variants in a single prompt, which guarantees they assess the same skill.
“Create a math exercise on fractions for a 6th-grade class. Generate three versions: (1) Guided version with hints and intermediate steps for struggling students. (2) Standard version aligned with the curriculum. (3) Challenge version with open-ended problems for advanced students. Each version must assess the same skill: adding fractions with different denominators.”
The last sentence is what holds everything together. Without it, the challenge version drifts toward a different skill—often multiplication or decimal fractions—leaving you with three assessments and no common rubric, unable to say whether the advanced student has mastered what you taught.
The same mechanism applies to specific needs, provided you name them. For multilingual learners, request simplified vocabulary, visual supports, and reformulated instructions; otherwise, the exercise assesses reading ability rather than the concept. For students with learning differences (such as dyslexia or dyscalculia), request short instructions, visual aids, and formatting tips. For gifted students, steer toward open problems, research tasks, and content creation: a student who writes three exercises for their peers works harder and more deeply than by solving ten additional problems of the same type.
Build a Bank Instead of Producing Ad Hoc
Generating an exercise the night before a lesson builds no capital; the next day, you start over. Instead, dedicate a full session to producing ten exercises per targeted skill. Sort them by difficulty from one to three stars, systematically add answer keys and grading criteria, and store everything in a folder or project shared with your team. By the end of a term, preparing a remediation session becomes a ten-minute selection process rather than a two-hour writing task, and a colleague taking over your class has access to the same materials.
Furthermore, you can recontextualize the same type of exercise at will to maintain interest without changing the difficulty: a sports context for a class that loves sports, a culinary context for working on proportions, an environmental context to cross-reference sustainable development education, or a tech context for digital enthusiasts. This variation costs one sentence in the prompt and visibly changes group engagement.
Verification and Validation
No generated exercise goes to students without a four-point check. First, solve it yourself: this is the only way to verify that the solution is correct and the reasoning valid, and errors lodge in answer keys far more often than in problem statements. Next, reread the instructions and ask if they are unambiguous for a student you aren’t guiding. Check alignment, meaning the exercise assesses the targeted skill and not a neighboring one—a proportionality problem buried in a long text primarily assesses reading. Finally, adjust the grading scale, which is often proposed according to a logic that isn’t yours and would penalize your students on points you don’t grade.
Your Turn
Your three versions.
- Pick a skill you assess regularly
- Generate the guided, standard, and challenge versions, naming the common skill
- Solve each version yourself
- Run the four-point check: solution, instructions, alignment, grading scale
- Note every adjustment: it signals a precision missing from your initial request
Official sources: Getting started with ChatGPT as an educator (OpenAI Help Center) · Prompting (ChatGPT documentation)
Key Takeaways
- Differentiate across three axes: content, process, and product
- The three-level simultaneous prompt is the most effective method
- Always specify the common skill for all three versions, or they will diverge
- Build an exercise bank rather than generating ad hoc
- Always solve the exercise yourself before distributing it