Following the publication of the systematic literature review we carried out last year, we wanted to create something that would make the findings more accessible in a cost-effective way. Tools like NotebookLM™ seemed like a promising option, allowing us to experiment with turning a dense academic paper into a short, engaging video without needing specialist production skills. And here is what NotebookLM™ was able to do for us.
As we worked through the process, however, we ran into a few challenges. One issue was oversimplification: the AI-generated summaries sometimes smoothed over important nuances, meaning we had to carefully check and edit the content to make sure the research was represented accurately. We also had to think about transparency and credit, particularly how to acknowledge the original work, the funder, and be clear about the role AI played in shaping the final output. Environmental considerations came up too. While creating a single video felt low-impact, we were conscious that AI tools still rely on energy-intensive systems, especially when used repeatedly. We weren’t able to find any research that compared the environmental cost of human-led video design to AI-led design.
We see this very much as part of our journey of understanding how we can make research accessible to a wider audience, and this experience showed us that AI tools can be genuinely useful for research communication, but only when combined with close human oversight and thoughtful reflection on their wider implications.
