Google has introduced a feature that can turn your research into an AI-generated podcast, featuring two “hosts” who discuss your findings. This experimental function is a part of NotebookLM, the AI note-taking app that Google unveiled last year. The AI hosts will “summarize your material, draw connections between topics, and engage in lighthearted banter. ” This feature builds on the existing capabilities of NotebookLM, enhancing how you interact with your notes, transcripts, and research documents. The app already employs Google’s Gemini AI model for summarizing research, and this new addition resembles an audio rendition of that process. Remarkably, Google genuinely means it when it claims that the AI hosts will “banter” with one another. I tested the Audio Overview with a sample notebook on the invention of the lightbulb, and the results were surprisingly realistic.

Over a 10-minute overview, the two hosts engaged in a jovial conversation about Thomas Edison not being the sole inventor of the lightbulb, concluding that “eventually, it’s really a story about teamwork making the dream work. ” The AI hosts could easily be mistaken for actual podcasters, as evidenced by their enthusiastic use of phrases like “bam!” in the middle of discussions and contemporary phrases such as “messy as heck. ” However, there were some oddities, like one host spelling out words and phrases like “P-L-U-S. ” Additionally, some dialogue lacked the natural flow of human conversation, with one host referring to platinum as “bling bling metal. ” While the feature enhances the engagement of learning about research, I wonder if the hosts would retain their playful tone when tackling more serious subjects, such as cancer or war. The conversation also included a fair amount of filler, potentially detracting from the clarity and efficiency of information delivery. Google acknowledges this in its announcement, noting that the feature offers “not a comprehensive or objective view of a topic, but simply a reflection” of your notes. Audio Overview comes with some limitations as well; the generation of a podcast-like discussion may take several minutes, and currently, it is only available in English. Like many AI tools, it isn’t always pinpoint accurate. If you're interested in trying this feature, simply open a notebook in NotebookLM. Then, select the Notebook guide located in the bottom-right corner of the screen and click Load under the “Audio Overview” section. I’m definitely planning to research a random topic just to see what the AI podcasters have to say!