The study, published in the journal AI & Society by Yasmin A. Carruthers and Johannes M. Heim, compared podcasts generated by NotebookLM with naturally produced human podcasts on the same topics, examining two hallmarks of human conversation: the timing of turn-taking and the way speakers adopt contextually appropriate conversational roles. The researchers argue that naturalness in dialogue requires far more than getting the timings right; it demands a contextually appropriate, addressee-oriented approach that responds to how a conversation actually unfolds.

Human turn-taking is a marvel of predictive coordination. Decades of research, beginning with the landmark work of Sacks, Schegloff and Jefferson in 1974, have established that conversation is governed by an organized system in which only one person typically speaks at a time, transitions are smooth, and gaps or overlaps are minimized. Listeners exploit morphosyntactic, semantic and prosodic cues to anticipate when a speaker’s turn will end, preparing their responses while the previous speaker is still talking. Average turn transitions fall in the range of roughly 200 to 250 milliseconds, far shorter than the approximately 600 milliseconds humans need to plan and produce a spoken response. This means that human interlocutors must be predicting, not merely reacting, and traces of this ability appear remarkably early in development, with infants showing sensitivity to conversational turn structure from birth.

Yet the “no-gap, no-overlap” principle is not absolute. Subsequent research has shown that overlaps can account for a substantial share of speaker transitions, and that brief overlaps often signal engagement rather than disruption. Backchannels, the brief vocal acknowledgments such as “mm-hm” or “right” that listeners produce without claiming the floor, add a dynamic and expressive quality to human dialogue. In podcast settings, hosts and guests also follow genre-specific social scripts: hosts introduce topics and then step back into a novice role to give guests a platform, while guests predominantly adopt expert stances. These role expectations, rooted in the broadcasting heritage of podcasting, make conversational behavior highly predictable despite frequent shifts between expert and novice positions.

Against this backdrop, the researchers annotated approximately ten-minute excerpts from two human podcasts and two NotebookLM-generated podcasts built from their transcripts, using the acoustic analysis software Praat. They coded speaker contributions, role shifts, turn durations, overlaps and transitions, achieving very high inter-annotator agreement. The human podcasts came from a series on productivity, featuring a recurring host and invited guests, while the AI versions were generated by NotebookLM using Gemini-based script writing and Google’s own text-to-speech voices, with no additional training data supplied.

On the surface, the AI podcasts behaved similarly to their human counterparts: questions structured the conversation, topics were expanded through follow-ups, and both hosts and guests provided feedback through backchannels. A statistical model confirmed that guests, whether human or artificial, produced significantly more expert contributions than hosts, in line with the conventions of the podcast genre. But the deeper measurements told a different story. Human turns were dramatically longer, averaging 22.3 seconds for hosts and 49.4 seconds for guests, compared with just 11.3 and 14.8 seconds for the AI agents. Human distributions of turn length were skewed toward long, developed contributions, while AI turns were short and symmetrically distributed.

The most striking divergence appeared in overlapping speech. Human guests overlapped for an average of more than four seconds and human hosts for over two seconds, whereas AI overlaps lasted mere milliseconds. AI backchannels rarely overlapped at all; instead, the current speaker paused, waited for the backchannel to be uttered, and then resumed, producing a mechanical politeness that no human interlocutor would display. The researchers interpret human overlaps as signals of engagement and emotional alignment, made possible by the fact that human listeners continuously predict where turns will end. NotebookLM, by contrast, sidesteps the problem of endpoint prediction entirely by pre-planning the entire dialogue script before converting it to speech, strictly enforcing the avoidance of gaps and overlaps in a way that humans never do.

The study also tested a newer NotebookLM feature that allows a human to join the AI conversation live, and the results were revealing. Every human intervention disrupted the pre-planned script: transitions slowed dramatically, overlaps vanished, audio glitches caused sudden voice switches, and the agents could only acknowledge the human’s contribution after long, silence-based delays before returning to their prepared narrative. Because spontaneous intervention requires resource-intensive, real-time context integration that pre-planning was designed to avoid, the hybrid dialogues lost the fluidity that made the fully scripted versions sound convincing. The authors conclude that pre-determining role behavior and turn-taking only goes so far, and that NotebookLM clearly hits its limits when responding to unpredictable input.

Taken together, the findings suggest that AI-generated dialogue can mimic individual features of natural conversation in isolation but fails to synthesize them into the dynamic negotiation of information and social relation that defines human talk. The AI agents spoke shorter, overlapped far less, and behaved more like two equally informed participants than like a host giving a guest a platform, breaking with genre conventions that listeners intuitively expect. Human conversation, the authors argue, is not optimized for seamless information exchange but for building common ground, an emergent and collaborative process shaped by mutual attention, verbal feedback and shared, growing context. Variable turn-taking timings and adaptive conversational role behavior remain unmatched human skills, and achieving genuine naturalness will require AI systems that can interpret social contexts and respond dynamically to the fluid negotiation of interlocutor relations, not merely imitate their surface elements.

Subject of Research: Turn-taking timings and conversational role behavior in AI-generated podcast dialogues compared with natural human conversation

Article Title: Not your average podcaster: turn-taking timings and conversational role behavior in AI-generated dialogues

Article References: Carruthers, Y. A., & Heim, J. M. (2026). Not your average podcaster: turn-taking timings and conversational role behavior in AI-generated dialogues. AI & SOCIETY. https://doi.org/10.1007/s00146-026-03365-3

Image Credits: AI Generated

DOI: 10.1007/s00146-026-03365-3

Keywords: AI-generated podcasts, NotebookLM, turn-taking, conversational roles, dialogue generation, backchannels, overlaps, knowledge asymmetry, common ground, pragmatics, human-machine interaction, speech technology

The study, led by Professor Philip Moons from the University of Leuven in Belgium, highlighted the innovative capabilities of a personalized AI research assistant, Google NotebookLM. As part of this project, researchers utilized this AI tool to produce podcasts that elucidate the findings of various scientific papers published in their journal. What was particularly surprising was the feedback from authors, with half of them believing that the podcasts were generated by human hosts rather than an AI system.

During the study, ten articles were selected across various formats, including original research articles, reviews, and patient perspectives. The authors of these articles were not informed about the AI-generated nature of the podcasts, which enabled the researchers to gather unbiased feedback. This methodological choice was vital to evaluate the tool’s effectiveness in creating engaging and trustworthy content without the influence of preconceived notions about artificial intelligence.

The quality of the AI-generated podcasts impressed many authors. They noted that the AI effectively encapsulated the essence of their research using simple, easy-to-digest language. The structure of the podcasts was praised for its coherence and for maintaining an appropriate length and depth, which made the complex scientific concepts more accessible to listeners. Several authors were astonished by the professionalism exhibited by the podcast hosts, with some expressing their belief that the speakers possessed a medical or nursing background.

While the podcasts were generally well-received, there were notable critiques regarding trustworthiness. Some authors pointed out that the AI’s American accent and stylistic choices sometimes detracted from the credibility of the content. Instances of using superlative language—such as "amazing," "groundbreaking," or "totally"—were highlighted, as they might have exaggerated the findings. Furthermore, a few inaccuracies were detected, including misrepresentations and mispronunciations, which underscored the need for thorough checks before publication.

The AI detection aspect of the study revealed both surprise and skepticism among the authors. When prompted about the AI-generated nature of the podcasts, half of the participants expressed astonishment, with emotional reactions ranging from shock to existential contemplation. The other half recognized the AI’s involvement, hinting at a learning curve associated with understanding how seamlessly AI can generate human-like content. This cognitive dissonance offers fascinating insights into current perceptions of AI technologies in scientific communication.

Authors strongly believed that the primary audience for these podcasts should be patients and the general public. The simplified and engaging format made it an ideal medium for conveying scientific knowledge to those who may not have the expertise to interpret dense academic literature. Additionally, some authors suggested that the podcasts could serve healthcare professionals by providing easy access to current research updates, fostering a culture of continuous learning within the field.

The feedback collected also illuminated the potential to tailor the podcasts for specific demographics, taking into account age, interests, and cultural backgrounds as AI technology evolves. Currently, capabilities to adjust the podcast’s voice or language remain limited; however, participants were optimistic that such features may become available in the future, further enhancing the accessibility of scientific information.

Professor Moons elucidated on the study, noting the promising accuracy of the AI-generated podcasts. While acknowledging this was only the beginning, he anticipates significant improvements in quality in the months ahead. The implications of being able to produce podcasts with minimal effort—possibly just by uploading an article—could revolutionize the way scientific knowledge is disseminated. This sustainable model would target those who do not engage with scientific journals, widening the reach of crucial research findings.

It’s essential, however, to stress that the rise of AI-generated content does not signal the end for human podcasters. There will continue to be a demand for human-created podcasts, as certain topics require nuanced understanding and interpretation that AI may not be able to fulfill effectively. The potential for hybrid models where human insights merge with AI capabilities offers a unique opportunity for enhancing science communication further.

The researchers have plans to delve deeper into the AI-generated podcast phenomenon. They aim to gauge responses from both patients and the broader public, seeking to understand their perceptions and experiences with these podcasts. Furthermore, they want to assess the feasibility of applying this technology to summarizing content from scientific conferences, thereby making complex discussions more digestible for those unable to attend in person.

As the landscape of scientific communication continues to evolve with the help of artificial intelligence, the implications of such innovations raise intriguing questions about the future of how research is shared and understood. By embracing these technologies while acknowledging their limitations, the scientific community can foster an environment where knowledge is democratized and made accessible to all, fulfilling the ultimate goal of education and public health.

In conclusion, the recent study not only demonstrates the exciting advancements being made in the field of AI and its applications in science communication but also presents a vital opportunity for enhancing public understanding of critical health information. As the journey of AI in podcast generation unfolds, researchers and audiences alike can anticipate a future where technology increasingly aids in bridging the gap between complex research and everyday understanding.

Subject of Research: AI-generated podcasts in science communication
Article Title: Artificial intelligence-generated podcasts open new doors to make science accessible
News Publication Date: 14 June 2024
Web References: European Journal of Cardiovascular Nursing
References: doi:10.1093/eurjcn/zvaf074
Image Credits: European Society of Cardiology

Keywords

Artificial intelligence, podcasts, science communication, healthcare, patient education, research dissemination.