The rise of AI-generated podcasts in science

AI-generated podcasts are making science more accessible. A new peer-reviewed study shows these tools could revolutionize how research reaches the public, with half of the participants unable to tell the podcasts weren’t human-made.

I don’t know about you, but when I listen to podcasts, the humanity and benevolence of the host are some of the most important things.

But when I heard that AI presenting is on the rise, I figured that perhaps I should have more of an open mind.

Researchers from KU Leuven used Google’s NotebookLM to generate AI podcasts explaining scientific articles from the European Journal of Cardiovascular Nursing.

The authors of 10 papers unknowingly reviewed podcasts generated by AI. They evaluated the content for engagement, trustworthiness, and whether they suspected AI involvement.

Feedback was gathered via a ChatGPT-assisted survey and follow-up interviews.

Prof. Philip Moons explained, “When I did a first test case with one of my own articles, I was flabbergasted by the high quality and how natural it sounded.”

Engagement and accessibility

In the findings, the authors found the podcasts clear, well-structured, and easy to understand. Interestingly, many believed the voices belonged to real professionals in nursing or medicine.

The conversational tone and balanced depth deemed them suitable for public consumption, showing the much-improved standard that seems to be happening with artificial hosts.

Most authors found the content trustworthy but flagged Americanized phrasing, minor inaccuracies, and pronunciation issues.

Half didn’t realize the podcasts were AI-generated; some were even “shocked” or had an “existential crisis” when told.

Implications and future use

The authors of the study stressed the need for clear AI labeling and referencing the original paper.

AI podcasts could help journals reach non-technical audiences, including patients and the general public. They may also support healthcare professionals in staying updated.

“This could be a sustainable model to get the message out to people who do not typically read scientific journals,” said Prof. Moons.