Your Learners are Using AI to Redesign Your Courses

What AI usage among learners in 2025 tell us about the current & future state of instructional design

Hey folks!

Picture this: a learner on a course you designed opens their laptop and types into ChatGPT: "I want to learn by teaching. Ask me questions about calculus so I can practice explaining the core concepts to you."

In essence, this learner has just become an instructional designer—identifying a gap in the learning experience and redesigning it using evidence-based pedagogical strategies.

This isn't cheating—it's actually something profound: a learner actively applying the protégé effect, one of the most powerful learning strategies in cognitive science, to redesign and augment an educational experience that, in theory, has been carefully crafted for them.

Two recent data releases—OpenAI's "Top 20 Chats for Finals" and Zao-Sanders' analysis of the most common AI use cases in the Harvard Business Review—reveal two critically important insights:

1. AI is now a go-to tool in the learning process for many learners: "Enhanced learning" is the #4 most common use case for generative AI.
   

2. Learners are using AI to fix the shortcomings of instructional design practices: the evidence is clear—learners are using AI to fix gaps in our design processes & practices.

In this week's blog post, I argue that instructional designers should study closely how learners are working with AI and evolve our role accordingly, from content creators to learning ecosystem architects.

“Enhanced learning” is now the #4 use case for generative AI (April, 2025)

Key insights:

• Learners are using AI to fix gaps in our design: poor scaffolding, missing emotional support, inadequate practice opportunities, and inaccessible content

• The sophistication of learner adaptations reveals that we're underestimating their capabilities and learning science knowledge

• The future ID role shifts from designing fixed experiences to creating adaptive learning ecosystems that learners can navigate and customise

• Instructional designers who embrace this evolution will become more valuable, not less

Let's dive in! 🚀

---

The Learner Behaviour Data That Should Change Everything

Let's start by exploring OpenAI's "Top 20 Chats for Finals" data, which reveals how learners worldwide actually engage with learning content. In a post in May 2025, OpenAI revealed not just how many learners are using AI to enhance the learning process—but also how they are using it.

The patterns in the data about the questions asked of AI by learners are striking. The most notable thing off-the-bat is that students are not using AI to "cheat"—they're using it to implement powerful instructional strategies to help them learn.

Here's the TLDR on the patterns I noticed:

Sophisticated Learning Strategies:

• "I want to learn by teaching. Ask me questions about [topic] so I can practice explaining the core concepts to you"

• "Identify and share the most important 20% of learnings from this topic that will help me understand 80% of it"

• "Create a practice quiz for me based on the material. Ask me each question one by one"

Emotional and Motivational Needs:

• "I'm not feeling it today. Help me understand this lecture knowing that's how I feel"

• "Motivate me"

Accessibility and Personalisation:

• "Can you take the following slides and help me learn the content in a faster and more interesting way?"

• "Decode this dense passage into language I can understand"

• "Create a game to help me [learning goals]"

Top 20 chats between OpenAI models & learners (May, 2025)

Process Support and Scaffolding:

• "Give me a step-by-step guide to help me finish [project]. Make the steps as small and achievable as possible"

• "Look for any rules and requirements in this assignment and make a checklist that's easy to understand"

• "Act as my public speaking coach and give me feedback to help me improve"
  

Critical Thinking and Academic Rigour:

• "I want to pressure test my thesis before I keep writing. Suggest the existing opposing viewpoints and any flaws in my logic"

• "I want to consider multiple perspectives. Find 3 experts with different points of view and compare their opinions"

Each of the most common prompts entered by learners doesn’t only dispel the “AI = cheating” myth — it also represents a learner actively rejecting our one-size-fits-all approach and creating the personalised, responsive, emotionally-aware learning experience they actually need.

What This Behaviour Reveals About Current Instructional Design

The data we are gathering about how our learners are using AI is uncomfortable but essential for our growth as a profession. Learner AI usage is essentially a real-time audit of our design decisions—and the results should concern every instructional designer.

"I use ChatGPT as a study guide to explain stuff that the course glosses over, which I then add to my notes. This helps me reinforce what I'm learning, and it's been hella useful so far." —Real student quote from Zao-Sanders' research

This quote should concern every instructional designer. A learner is saying our course "glosses over" important concepts, and they're turning to AI to fill the gaps we've left. Here's what the patterns reveal:

We're Designing for Institutional Compliance, Not Learning Effectiveness

When learners need AI to "make a checklist that's easy to understand" from our assignment instructions, it reveals that we're designing to meet organizational requirements rather than support learner success. We're optimizing for administrative clarity rather than learning clarity.

The evidence: Learners consistently ask AI to translate our instructions into actionable steps, suggesting our task design prioritizes comprehensive coverage over usability.

Our Content Delivery Ignores How Brains Actually Learn

Learners asking AI to "explain [topic] to me in 3 different levels of understanding" or "turn this into flashcard-style questions" are essentially redesigning our instructional experiences in real-time. This exposes that our materials are:

• Too one-dimensional (single explanation style)

• Poorly formatted for retrieval practice

• Missing scaffolding for different cognitive loads

• Ignoring spacing effects and interleaving that cognitive science tells us works

The evidence: Learners systematically request multiple representations, active recall formats, and spaced practice—fundamental learning science principles we often ignore in our designs.

We've Underserved the Emotional Dimension of Learning

The popularity of prompts like "I'm not feeling it today. Help me understand this lecture knowing that's how I feel" and "Motivate me" reveals a massive gap in our design thinking. We design as if learning is purely cognitive when research clearly shows emotional state directly impacts cognitive capacity.

The evidence: Learners are seeking emotional metacognition and motivational support that our learning experiences simply don't provide, despite decades of research on motivation and engagement.

We're Underestimating Learner Sophistication and Agency

When learners ask AI to "identify the most important 20% of learnings... that will help me understand 80% of it," they're applying the Pareto principle to learning. When they request "pressure test my thesis... suggest existing opposing viewpoints," they're seeking rigorous critical thinking support.

The evidence: Learners demonstrate sophisticated understanding of learning strategies, prioritisation frameworks, and intellectual rigor that our designs often treat as advanced concepts rather than foundational needs.

Our Feedback and Practice Systems Are Fundamentally Broken

Learners asking AI to "act as my professor and grade this draft" or "create a practice quiz... ask me each question one by one" reveals that our assessment and practice systems fail to support learning.

The evidence: Learners crave low-stakes practice with immediate feedback, but our systems typically provide high-stakes evaluation with delayed feedback—exactly the opposite of what learning science recommends.

How Learners Are Fixing Our Design Flaws (And What This Teaches Us)

Let's examine specific examples of how learners are adapting our carefully crafted experiences—and what their adaptations reveal about our design shortcomings:

When we provide: Dense academic readings with complex language
Learners adapt by asking AI: "Decode this dense passage into language I can understand"
What this teaches us: Our content often prioritises academic sophistication over accessibility, creating barriers rather than bridges to understanding

When we provide: Overwhelming project requirements buried in lengthy documents
Learners adapt by asking AI: "Look for any rules and requirements in this assignment and make a checklist that's easy to understand"
What this teaches us: We're not making success criteria transparent and actionable—we're hiding them in administrative prose

When we provide: Static content delivery methods
Learners adapt by asking AI: "Can you take the following slides and help me learn the content in a faster and more interesting way?"
What this teaches us: Our delivery methods often ignore engagement principles and different learning preferences

When we provide: High-stakes final assessments with no practice opportunities
Learners adapt by asking AI: "Create a practice quiz for me based on the material. Ask me each question one by one"
What this teaches us: We're not providing enough low-stakes practice with immediate feedback—the foundation of skill development

When we provide: Limited perspectives on complex topics
Learners adapt by asking AI: "I want to consider multiple perspectives. Find 3 experts with different points of view and compare their opinions"
What this teaches us: We're not scaffolding critical thinking and multiple perspective analysis effectively

When we provide: Generic motivational content that ignores emotional states
Learners adapt by asking AI: "I'm not feeling it today. Help me understand this lecture knowing that's how I feel"
What this teaches us: Our emotional design completely ignores the reality of how emotional state affects learning capacity

When we provide: Abstract concepts without sufficient scaffolding
Learners adapt by asking AI: "Give me a step-by-step guide to help me finish [project]. Make the steps as small and achievable as possible"
What this teaches us: We're not breaking down complex tasks into manageable components that support skill building

What Learners Are Really Demonstrating (That Should Inspire Us)

Here's the part that should both humble and inspire us as instructional designers: learners aren't using AI to avoid learning—they're using it to create more effective learning experiences than we've designed for them.

They're using advanced learning strategies we don't teach. The learner who asks "I want to learn by teaching. Ask me questions about [topic] so I can practice explaining the core concepts to you" is using the protégé effect—one of the most powerful learning strategies in cognitive science.

They're applying sophisticated prioritisation frameworks. When a learner asks "Identify and share the most important 20% of learnings from this topic that will help me understand 80% of it," they're applying the Pareto principle to learning—something we rarely teach explicitly.

They're seeking emotional metacognition. The learner who says "I'm not feeling it today. Help me understand this lecture knowing that's how I feel" is demonstrating sophisticated awareness of how emotional state affects learning capacity.

They're demanding accessibility without compromising rigor. When learners ask AI to "decode this dense passage into language I can understand" while also asking to "pressure test my thesis... suggest existing opposing viewpoints and any flaws in my logic," they want both accessibility and intellectual challenge.

They're creating scaffolded practice environments. Learners asking for "practice quiz... ask me each question one by one" or "step-by-step guide... make the steps as small and achievable as possible" are building the deliberate practice environments we often fail to provide.

The Future of Instructional Design: From Content Creators to Learning Ecosystem Architects

Learner behaviour is pointing toward a fundamental evolution in our role as instructional designers. The future belongs to designers who can create adaptive learning ecosystems rather than fixed learning experiences.

This shift begins with reconceptualising our primary function. Instead of crafting the "perfect" explanation or linear sequence, we need to design modular learning components that learners can combine and adapt based on their needs. When learners ask AI to "identify the most important 20% of learnings that will help me understand 80% of it," they're asking for content restructured by learning priorities rather than academic comprehensiveness. We're evolving from content creators to experience architects who design interconnected learning nodes navigable through multiple pathways.

The assessment dimension is equally profound. When learners consistently ask AI to "create a practice quiz... ask me each question one by one" or "act as my professor and grade this draft," they reveal that our high-stakes, infrequent evaluation approach fundamentally misunderstands how learning happens. We need to become practice environment builders who embed continuous feedback loops throughout the journey, thinking like game designers rather than test writers—creating systems where learners naturally want to improve through immediate feedback and failure without penalty.

Perhaps most significantly, we must evolve how we approach the emotional dimension of learning. Prompts like, "I'm not feeling it today. Help me understand this lecture knowing that's how I feel", reveal that learners intuitively understand what we often ignore: emotional state directly impacts cognitive capacity. We need to evolve from adding inspirational quotes to building psychological safety, adaptive support, and emotional awareness into learning architecture itself—designing for emotional states as explicitly as we design for cognitive load.

The metacognitive dimension represents our profession's most exciting opportunity. Learners are discovering sophisticated learning strategies through AI—the protégé effect, Pareto principle application, emotional regulation, spaced practice—that we often assume they possess or treat as advanced concepts. Our role shifts from subject matter organisers to metacognitive skill teachers who make learning strategies as explicit and teachable as content knowledge.

These emerging learning ecosystems look fundamentally different: multiple entry points rather than predetermined sequences, integrated practice and feedback rather than front-loaded content, scaffolding for emotional dimensions rather than assuming neutrality. Most importantly, they embrace learner agency as a design principle rather than viewing it as a threat to instructional control.

This transformation makes instructional designers more valuable, not less relevant. While AI can generate content and provide personalised tutoring, it cannot make complex architectural decisions about learning ecosystem design, emotional scaffolding, and metacognitive skill development. The future belongs to instructional designers who can orchestrate human-AI collaborations to create learning experiences that are both more effective and more humane than anything either humans or AI could create alone.

Conclusion: A New Professional Opportunity

The most successful instructional designers of 2025 and beyond won't be those who resist AI or those who blindly embrace it. The winners will be those who study what learner AI behaviour teaches us about effective learning design—and who use those insights to create more responsive, human-centered learning ecosystems.

This evolution makes instructional designers more valuable, not less. While AI can generate content and provide personalised support, it cannot make the complex design decisions about learning ecosystem architecture, emotional scaffolding, and metacognitive skill development that learner behaviour reveals are essential.

The revolution is here—and it's being led by learners showing us what effective learning design could look like. At the center of it all? A quiet conversation between a learner and an AI, revealing what we got right, what we got wrong, and what we could do better.

The future of instructional design isn't about competing with AI—it's about learning from what learners create when they have access to responsive, personalized, emotionally-intelligent learning support.

Happy innovating!

Phil 👋

PS: Want to develop the skills needed for this evolving instructional design landscape? Join me in exploring how to design learning ecosystems that work with AI rather than against it on my AI Learning Design Bootcamp.