How Do You Make a Podcast With NotebookLM?

How Do You Make a Podcast With NotebookLM?

You upload a few documents, click one button, and a couple of minutes later two AI hosts are discussing your material as if they had prepared for it.

That is the thing people mean when they say "the NotebookLM podcast," and it is genuinely the most impressive party trick in the product.

It is also the thing with the worst name mismatch in Google's lineup, in two directions at once. Google has never called it a podcast. And Google no longer calls the product NotebookLM either.

Neither of those changes what you click. But both of them change what you should search for when something goes wrong, which is why this guide starts there before getting to the steps.

What is the short answer?

Open a notebook, add your sources, and in the Studio panel select Audio Overview. Google's own instruction is exactly that: "In the 'Studio' panel, select 'Audio Overview' to generate."

Generation runs in the background, so you can keep working in the notebook while it goes. Google says it "may take a couple of minutes to generate the Audio Overview."

One prerequisite that catches people out on shared notebooks: you need edit access to generate or delete an Audio Overview. Read-only access lets you listen to one somebody else made, and nothing more.

Publishing content that AI assistants should be citing? Get a free SEO and AI visibility audit and see where you currently stand.

Two name changes worth knowing about

This matters practically, not pedantically. If you search Google's help center for "NotebookLM podcast," you are searching for two things Google does not call by those names, and the results get noticeably worse for it.

It is not called a podcast

Google's term is Audio Overview, and the definition on its help page is specific: "Audio Overviews are deep-dive discussions between AI hosts that provide in-depth summaries of the key topics in your uploaded sources."

The same page adds a framing that is easy to miss and useful to know: they are "designed to be an objective reflection of your source content, rather than subjective opinions from the AI hosts." The hosts sound like commentators. Google's stated intent is that they summarize rather than opine.

The word "podcast" does not appear on any of Google's Audio Overview pages.

It is entirely user vocabulary, which is a reasonable thing for it to be — the output does sound like a two-host show — but it is not a search term that Google's own documentation will match.

It is not called NotebookLM

The help center for the product now lives at a gemininotebook address and is titled Gemini Notebook Help.

The product's own site describes itself as "Gemini Notebook, the AI research tool and thinking partner," and the help article's opening line reads: "Gemini Notebook is an AI-powered research assistant designed to help you refine and organize your ideas."

The old name survives only in plumbing — in URLs, in tracking parameters, and in the support address [email protected], which is still the one the FAQ gives you. In actual product copy it is gone.

What Google does not publish anywhere I could find is an announcement of the rename. There is no "formerly NotebookLM" banner and no dated statement on the help pages.

Your notebooks, sources and saved notes are all still there; the name over the door simply changed without a notice on it.

Wondering whether your own pages are being summarized accurately by AI tools? Request a free audit and find out.

Generating one, step by step

1. Open an existing notebook or create a new one, and upload your sources. The notebook-level list Google publishes is PDFs, websites, YouTube videos, audio files, Google Docs and Google Slides. Note that this is the list for sources generally — Google does not publish a separate list of source types that Audio Overview specifically accepts.
2. Check you have edit access if the notebook belongs to someone else.
3. In the Studio panel, select Audio Overview.
4. In the generation panel, set the format, the language, and the length, and optionally type a prompt to steer it.
5. Generate, and leave it running. It goes in the background and takes a couple of minutes.
6. If you would rather not wait at all, the web version offers a "generate later" path. Google is candid about the trade: "Generating later may take a couple hours but you will be notified if you have notifications enabled."

To come back to one you already made, open the notebook and find it in Studio, then select Load.

If you used a prompt and have forgotten what you asked for, the three-dot menu has a View custom prompt option, which is more useful than it sounds once you have made a dozen of these.

If you are doing this on a phone, temper your expectations: Google notes only that "the Gemini Notebook mobile app may have limitations to this feature," without saying which limitations.

The four formats, and when each is right

The format picker is where most of the quality comes from, and it is the control people most often leave on its default.

• Deep Dive is the default, and it is the familiar two-host conversation most people have heard.
• The Brief is the one to reach for when you do not want a show at all. Google describes it as: "A single speaker delivers the key takeaways... in under two minutes." That is the only duration figure Google publishes for any format.
• The Critique turns the hosts into reviewers — "Two hosts provide a constructive evaluation" — which is a genuinely different use, closer to feedback on a draft than a summary of it.
• The Debate sets them against each other, with "two hosts engage in a formal, back-and-forth debate." Useful for material with real disagreement in it, and misleading for material without any.

Alongside the format you get a length control offering Shorter, Default or Longer, with one caveat attached in Google's own text: "(English Only)." So on a non-English Audio Overview the length choice is not available to you.

The prompt field is the highest-leverage control of the lot. Google's description is that you can "input a prompt to provide instructions to focus on specific topics or adjust the expertise level." Telling it to assume the listener already knows the basics is usually worth more than any format change.

Interactive mode: joining the conversation

You can talk to the hosts rather than only listening. Google's own words: "You can join a conversation and interact with the AI hosts in Audio Overview."

The flow is: create a new Audio Overview, select Interactive mode, then while it plays, select Join.

Three restrictions are documented and all three are easy to trip over. "The interactive mode experience is currently available only in English." It works only on newly generated Audio Overviews, so you cannot switch it on for one you made last week.

And it does not travel: "Other users won't be able to interact with your Audio Overview through the shared link."

Google also warns about the seams, which is fair of it: "You might experience some audio glitches like random speaker switches, third voice, or voice glitches," plus a slight delay when you join.

Building a content library you want AI tools to pull from? Start with a free audit and see what is reachable today.

Sharing it, and getting the file out

In the audio player there is a Download option and a Share option.

Download is where Google's documentation stops unhelpfully short: it says to "select Download" and never states what file format you get. No format is named on any of its Audio Overview pages, so plan on finding out by doing it.

Sharing is tied to the notebook rather than to the audio. You have to "confirm that the notebook is shared with the recipient or is made public," meaning open to "anyone with a link." There is an account-type restriction on the public route: "Public notebook sharing is only enabled for consumer accounts. It's currently disabled for Workspace Enterprise or Education accounts."

Two further traps. Deleting the audio breaks any link you already sent — "once an audio is deleted, any previously generated share link will no longer work." And if a Google Play Books ebook is one of your sources, downloading may be blocked outright, "depending on the publisher's content restrictions."

What the limits actually are

If you came looking for "how many can I make per day," Google does not publish that number for any tier.

What it publishes instead is a compute-based model, and the honest summary is that it is deliberately imprecise: "The limits factor the complexity of your prompt, the models and features you use, the length of your chat and specific features."

The tiers are expressed as multiples rather than counts. Without a plan you get standard limits; Google AI Plus is "two times higher than standard limits"; AI Pro is "four times higher"; and AI Ultra is "5x or 20x higher than AI Pro depending on your subscription." Google appends its own disclaimer: "Gemini Notebook limits may change."

Here is a contradiction worth knowing about before you plan around either number, because both pages are currently live. The usage page says "the quota refreshes every 5 hours until you reach your weekly limit." The plans page says "if your plan is on daily quotas, these reset after 24 hours."

The same page adds: "Monthly quotas reset after 30 days." A five-hour refresh against a weekly cap and a daily reset against a monthly cap are not the same scheme, and Google does not reconcile them.

One genuinely good piece of news in the small print: the artifacts the product sometimes makes for you unprompted do not count against you.

Google states that an automatically created report, flashcard set, infographic, slide deck, audio overview or video overview is "generated once and do not contribute to limits."

The per-source ceilings are published and generous: "The current limit is 500,000 words per source or up to 200MB for local uploads. There's no page limit."

Where it falls down

Google is upfront about the core caveat, and it belongs in any workflow you build on this: "Audio Overviews, including voices, are AI-generated and may contain inaccuracies or audio glitches." The grounding in your own sources makes that much less likely than with an ungrounded chatbot, but "less likely" is not "checked."

The named absences are the bigger practical problem. Google does not publish the audio file format, any maximum audio length, any minute figure for Default or Longer, any way to choose or name the host voices, or any option to export a transcript.

The transcript gap is the one I would most like closed, because an Audio Overview you cannot search is an Audio Overview you will not revisit.

Want to know which of your pages AI research tools can actually reach? Get a free audit and find out what is visible.

Final thoughts

The workflow is three clicks and the quality is largely decided by two controls most people never touch: the format and the prompt.

Pick The Brief when you want information and Deep Dive when you want company, and tell it what the listener already knows.

The naming is the thing to carry away. Search for Audio Overview rather than podcast, and for Gemini Notebook rather than NotebookLM, and Google's documentation becomes considerably more useful than it first appears.

For the wider picture of what the product does beyond audio — the sourcing model, the plan limits and where the rename leaves you — our guide to what Google NotebookLM is and how it works now covers the ground this article deliberately skipped.