For many of us, recorded lecture videos once felt like the holy grail of flexible teaching. They were easy to deploy, scalable for large cohorts and allowed students to engage with content on their own terms. But over time, the cracks began to show.
At UTS, learning analytics painted a stark picture. Early‑semester engagement with online lecture videos typically sat around 40%, but by mid‑ to late‑semester this figure often collapsed to closer to 20%. Even when students clicked on videos, average completion hovered around 40–45%. Lecture‑linked exam questions reflected this pattern, with fewer than 30% of students answering correctly.
This wasn’t just a data problem. It was a learning problem.
When passive content stops working
Decades of research now show that passive content consumption alone is insufficient for sustained learning, particularly in large or online cohorts (Freeman et al., 2014; Chi & Wylie, 2014). Watching long videos, skimming PDFs, or re‑reading slides tends to promote surface‑level engagement rather than understanding.
This challenge is amplified in disciplines like motor learning and skill acquisition, where conceptual understanding improves through application, feedback, and reflection, not memorisation. When learning design prioritises information delivery over interaction, students disengage, and often disappear.
The solution is not just “better videos” – it’s better learning design.
Most educators know that active learning works – structured practice, retrieval and application lead to stronger learning outcomes and lower failure rates. Yet redesigning online content to support this kind of engagement is often seen as technically complex, time‑consuming and difficult to scale, especially without dedicated support. This is where GenAI offers a genuinely useful intervention: not as a replacement for teaching, but as a design accelerator.
Rethinking the online lecture
Rather than asking AI to teach students, the approach showcased in my recent workshop reframes its role: GenAI supporting educators to design richer learning experiences, while academics retain full control.
Traditional recorded lectures were replaced with interactive Canvas pages built from existing slides and narration transcripts. Using carefully crafted prompts, GenAI transformed this material into structured, accessible HTML pages that integrated:
- Short, segmented explanations
- Visual diagrams and summaries
- Embedded low‑stakes quizzes
- Applied, case‑based learning tasks
Crucially, this was a human‑in‑the‑loop workflow. AI handled the initial build and structuring, but all content was reviewed, refined, contextualised and quality‑assured by the academic. Disciplinary accuracy, learning outcomesand relevance to students’ applied contexts remained non‑negotiable.
Interaction over content
A key ingredient in the redesign was the use of H5P interactive activities embedded directly into Canvas. These weren’t decorative add‑ons. They served clear pedagogical purposes:
- Supporting retrieval practice
- Making understanding (and misunderstanding) visible
- Encouraging persistence through content
- Scaling interaction without increasing staff workload
Research consistently shows that interactive and constructive learning modes lead to stronger outcomes than passive acquisition, particularly in online environments where engagement can easily fade.
What happened to engagement?
Early results are promising.
While interactive pages often received slightly fewer initial clicks than lecture videos, completion rates were far higher, around 78% compared to ~40% for videos. Once students began engaging, they were far more likely to persist through the content, rather than dropping off halfway through.
Formal evaluation is ongoing, with direct comparisons planned between video‑based delivery and GenAI‑assisted interactive modules, examining engagement, completion and exam performance. But the signal is already clear: active learning scales when design friction is reduced.
Why this matters beyond one subject
This approach isn’t about replacing lecturers with AI. It reflects what emerging scholarship continues to emphasise: AI is most effective when it supports educators’ design and evaluative work, not when it replaces it.
The workflow shared in this workshop is adaptable, repeatable, and discipline‑agnostic. It offers educators a practical pathway to redesign online learning in ways that are evidence‑informed, scalable, and sustainable, without requiring specialist technical skills.
If you’ve ever looked at your analytics and wondered why students disengage despite your best intentions, this work offers a potentially compelling alternative.
Leveraging GenAI in your teaching
Watch William’s full GenAI Workshop presentation video (50 mins; passcode: .@*njj0L)
Join us for future explorations of how you can leverage GenAI in your teaching: