Though Google’s AI-powered research assistant, NotebookLM, offers a lot, its Audio Overviews feature is one of the most standout additions. If you haven’t NotebookLM yet, you can convert your sources into a podcast hosted by two virtual hosts, which Google refers to as Audio Overviews. They’re engaging, fun to listen to, and have a hint of wit and humor in them, which helps turn the most boring topic into something you won’t dread studying.

I’ve been using NotebookLM since it launched as an experimental project, and the Audio Overviews feature since it was introduced in September 2024. When NotebookLM first introduced this feature, it used Gemini 1.5 Pro’s capabilities. Interestingly, even though it used Gemini’s capabilities, the feature was strangely not available directly in the Gemini interface. It was restricted to NotebookLM, until March 2025, when Google finally made Audio Overviews available in Gemini.

Though it’s been a couple of weeks since, I never tried it and stuck to using the feature in NotebookLM… until recently. And surprisingly, they take the already impressive Audio Overviews to the next level.

Gemini's Audio Overviews can pretty much do everything NotebookLM's can

Close match, different vibe

Let me be clear here — the core concept of Audio Overviews is the same in both NotebookLM and Gemini. Both NotebookLM and Gemini’s Audio Overviews are powered by Gemini, and both generate podcast-style audio summaries with two AI hosts. But the experience of using Audio Overviews in Gemini is slightly different, and in some ways, better.

To use any feature in NotebookLM, you need to create a notebook. Once you create a notebook and add sources to it, you can click the Generate button under Audio Overview to generate an AI podcast of your sources. As far as Audio Overviews are concerned, that’s all you can really do in NotebookLM — any Audio Overviews you create in NotebookLM are generated only using the sources you add to your notebook.

Gemini takes this up a notch. While Gemini’s Audio Overviews can also transform documents you upload into audio summaries, it can additionally convert Deep Research reports into podcast-style audio discussions too. Ultimately, Gemini’s Audio Overviews aren’t just restricted to the sources you provide. Though this means there’s a greater chance of hallucination compared to NotebookLM, I’ve found that the added flexibility makes it worth it, especially when you want a broader overview of a topic that goes beyond just your own materials.

I’ve also noticed Gemini’s Audio Overviews are much shorter than the ones you’ll get with NotebookLM, even when you use the same sources in both tools. So, I’ve been using Gemini's Audio Overviews when I want a quick and snappy overview, and NotebookLM when I want a more in-depth and longer overview.

Gemini lets you convert Deep Research reports into Audio Overviews

Skip the export step

What really tempted me to use Audio Overviews in Gemini is its integration with the Deep Research feature. If you haven’t used Gemini’s Deep Research before, it does exactly what it sounds like. You give Gemini the details of a topic you’d like to research, it formulates a research plan, analyzes the results, and then creates a report. For instance, in this example, I asked Gemini to use Deep Research to conduct research about XDA.

It came up with an 8-step plan, beginning with researching the foundational history and mission of the site, and ending with the current relevance of XDA. Deep Research evaluates dozens of websites, including community forums like Reddit, which really helps it paint a well-rounded picture of the topic. While Gemini is conducting its research, it even shows its entire thinking process, the research approach it’s taking, and where it’s gathering its information from. For example, in this case, it researched 104 websites, including Reddit, MuckRack, YouTube, and more.

Now, here’s the thing: the reports it produces are, well, deep. The final report Gemini produced in this example was nineteen pages long! That’s where Gemini’s Audio Overviews come in. Instead of reading through pages and pages of text, you can convert the report into a podcast right within Gemini.

Before Google added Audio Overviews in the Gemini interface, I’d download the Deep Research report, create a new notebook, upload it as a source, and then create an Audio Overview. Now, with Audio Overviews built directly into Gemini, that extra hassle is gone. Once your report is created, all you need to do is click the Create button and select Audio Overview from the dropdown. I’ve used this approach when I wanted to go beyond my sources without being limited to what I’ve uploaded to my notebooks.

Gemini’s Audio Overviews still aren’t perfect, though

Better in some ways, worse in others

I already find NotebookLM’s Audio Overviews impressive, so even before trying out Gemini’s, I knew I wouldn’t be disappointed. The integration with the Deep Research feature, and the fact that it can also generate Audio Overviews using sources uploaded like NotebookLM, made it even more appealing. But that doesn’t mean they’re instantly better than NotebookLM’s. Gemini’s Audio Overviews lack a lot of features that NotebookLM’s already has.

For instance, you can use the interactive mode in NotebookLM to jump into the podcast and ask questions or share your own opinions, which I find great for studying. Gemini’s, on the other hand, are currently a one-way listening experience. Additionally, like I mentioned above, Gemini’s Audio Overviews are generally much shorter than NotebookLM’s.

The latter even lets you pick between shorter and longer Audio Overviews now, but Gemini offers no control over the length of the audio. For example, the Audio Overview Gemini generated using the 19-page Deep Research report was only six minutes long.

You're missing out if you haven't tried Audio Overviews in Gemini

If you’ve been using NotebookLM’s Audio Overviews and haven’t tried Gemini’s yet, thinking they’re just the same (as I did for a long time), you’re missing out, especially when it comes to Deep Research.