Google turned my book into a podcast 😳

You’ve got a book, report, or research folder you want people to understand. They’ve got a calendar that suggests otherwise. Could turning it into a podcast help?

I tried it with The Comeback Code using Google’s NotebookLM, a tool that answers questions and creates useful formats from sources you choose.

Google made a podcast about my book, and I didn’t record a word. No studio, no microphone, no calendar negotiation. Apparently, the easiest guest to book is your own PDF.

The interesting question wasn’t whether it could make convincing audio. It was whether someone could listen and walk away understanding the ideas I actually wrote.

My book became a 21-minute conversation

I uploaded The Comeback Code and requested an Audio Overview: a generated spoken discussion based on the material you provide. Two remarkably energetic artificial intelligence (AI) hosts discussed ideas I’d spent years developing. The result ran 21 minutes.

Hearing familiar ideas reorganized and explained from another angle was the interesting part. I haven’t compared the audio with the book passage by passage, so I can’t tell you how faithfully it captured those ideas.

I’d start with the book’s five comeback stages. Does the discussion explain why each stage matters? Does it preserve their differences, or make distinct ideas sound interchangeable?

That’s a better test than “Did it sound like a podcast?” You want another way into the material, not a slightly different version of what it means.

The same test applies to your report, customer interviews, or training material. Before you give the hosts anything to discuss, it helps to know what you want someone to understand.

Choose the question before you choose the files

For a first experiment with my book, I’d ask: “What should a business leader who hasn’t read it understand about the five comeback stages?”

That question tells me what belongs in the notebook. The book belongs. A folder of loosely related leadership articles probably doesn’t.

NotebookLM can work with a book, report, meeting transcript, presentation, website, or notes. Unlike an open-ended conversation with a general AI assistant, this work centers on the collection you choose.

Think of a fast research partner reading your packet. It hasn’t necessarily read the right packet. That part is still your job.

You might use it to prepare for a meeting, compare competing ideas, find strategy gaps, extract customer lessons, or create training. Whatever the job, give it a notebook of its own.

“Leadership Stuff” is a storage label. “Prepare the executive team to decide whether we enter the healthcare market” gives the work a purpose.

For that healthcare decision, I’d include the current plan, customer research, financial assumptions, and the meeting transcript where people debated it. The company picnic menu can sit this one out, unless the potato salad contains a market signal.

You’re building a useful packet, not a large one. Each document should help answer the question, come from somewhere you trust, and be current enough for the decision. Another copy of the same claim doesn’t necessarily add anything.

Clear filenames help, too. FINAL, FINAL-2, and FINAL-REAL aren’t a filing system. They’re a short history of disappointment.

With sensitive material, permission to read a document isn’t automatically permission to upload it. Your organization’s rules and the tool’s current privacy, data-handling, and account terms matter here. Include only information you’re authorized to process, and no more than the job needs.

Find the weak assumption before it gets a microphone

Once you’ve loaded the sources, it’s tempting to go straight to audio. I’d spend a little time in the notebook’s chat first, asking what the argument rests on.

For my book, I’d want to see how the selected passages explain the stages and how the stories support those explanations. A story can make an idea memorable without proving every conclusion someone draws from it.

Here’s a prompt you can paste into the chat:

Work from these documents.
Here’s what I need to decide or explain: [your question].
This is for: [your audience].
Explain the main argument and show the passages supporting its most important claims. Point out missing information, disagreements, and assumptions being treated as facts.
Separate what the documents say from your interpretation or advice. Don’t fill gaps with invented facts. If the evidence is weak, say so.
Before we make an audio overview, what should I check in the originals?

The answer should give you citations, references pointing back to supporting passages. Opening those passages is where the useful work begins. Asking for evidence doesn’t guarantee the tool has understood it correctly.

Here’s the distinction that matters: a citation tells you where a claim came from, not whether it’s true.

Imagine the healthcare plan, presentation, and meeting notes all repeat the same demand estimate. Three documents repeating one untested assumption don’t give you three independent reasons to believe it.

If you trace that estimate back, you might discover it all came from the same initial guess. The notebook could accurately summarize your company’s confidence without finding any new evidence for it.

That’s a useful discovery, not a failed summary. It also changes what you ask the hosts to do.

Instead of “Explain why we should enter healthcare,” you’d ask them to explain the case and identify the unverified demand estimate. You’ve stopped an assumption from becoming a recommendation through repetition.

If the answer still feels too tidy, I’d follow that thread: which three claims carry the argument, and which have genuinely separate support? Disagreement between sources may tell you more than another paragraph of agreement.

Perhaps the demand estimate changed between an earlier plan and a later one. Asking what changed—and which questions remain unanswered—helps you see whether the evidence improved or just the forecast.

You could also ask for the strongest evidence against the recommendation, or what a skeptical CFO would challenge. You don’t need to run every question. Follow the one that could change the decision.

Now give the audio a listener and a purpose

Once you understand what the sources support, you can give the hosts a better job than “Make this interesting.”

For my book, I might use this instruction:

Create an engaging overview for business leaders who have not read The Comeback Code. Focus on the five comeback stages, explain why each stage matters, and use stories from the book when they clarify the lesson. Keep the tone practical rather than promotional. Do not introduce claims that are not supported by the selected sources.

I’m telling it who’s listening and what they should understand—not just asking it to sound interesting. Select the sources you want the hosts to use, then add your instructions if your account offers that option.

This is also where your earlier discovery belongs. If the healthcare estimate is unverified, tell the hosts to preserve that distinction. Otherwise, the conversation could turn “we expect demand” into “customers are waiting.”

When I listen, that’s the kind of change I’m listening for. For the book, I’d compare the explanations of the five stages with their original passages. I’d want the examples to clarify those explanations, not pull them toward a different meaning.

A careful statement becoming an absolute deserves another look. AI hosts enjoy certainty almost as much as human podcast hosts do.

If that happens, narrower sources or a revised instruction can give the next version a better chance. You don’t have to keep a mediocre result just because it arrived wearing headphones.

Use audio for understanding, writing for inspection

A 21-minute discussion gives someone another way into a book. They might listen before reading, revisit ideas while walking or driving, or arrive better prepared for a team discussion.

But our healthcare team needs more than a good introduction. They need the demand estimate and its supporting evidence where everyone can see and challenge them. I’d use a written briefing for that, with audio as an optional introduction.

The distinction is simple: listening helps you get oriented; writing lets you inspect. Nobody wants to scrub through a podcast to find a contractual commitment.

The same sources can support an FAQ for recurring questions, a study guide for learning, or a timeline of events and decisions. A mind map—a visual outline of connected ideas—can help you see relationships across the material.

If you want presentation support, look in Studio, the area where NotebookLM creates these outputs. The available options depend on your account.

There’s no prize for pressing every button. Choose the format that makes the next useful thing easier for your listener or reader.

Two questions before you share it

The healthcare example gives us the first question: could someone act on a wrong claim in this recording?

If that demand estimate sounds settled, someone might approve spending against it. That’s why I’d return to the original before sharing—not because every sentence needs a committee.

My line is any claim that could affect money, people, policy, reputation, health, or a customer promise. Names, numbers, quotations, dates, and commitments deserve particular care, as do surprising or unfamiliar claims.

The important distinction is what the source actually says versus what the AI has interpreted or advised. Confidence is a presentation style, not a warranty.

For my book, the consequence is different, but the question still works. Could a listener leave with a meaning the original doesn’t support? A pleasant discussion wouldn’t make that a good introduction to the book.

The second question is: are you allowed to share everything in the recording?

A meeting transcript might be approved for internal use while containing private customer details that don’t belong in a public episode. Removing private or unnecessary information matters even when every word is accurate.

Owning access to the source doesn’t settle publishing rights, either. For a public release, you need rights to the material and permission under the product’s current terms, including any publishing restrictions.

The audience matters here, too. An internal team may understand the context that outside listeners won’t, and the tone needs to fit them. Depending on where you’re sharing it, AI involvement may also call for disclosure or further review.

Those two questions keep the review tied to real consequences. An internally useful listening aid isn’t automatically a publication-ready podcast.

Try one document you already know

Give yourself five minutes to set up the experiment. Audio generation and listening may take longer.

Start with one familiar chapter, report, or research packet you’re permitted to use. Put it in a notebook around one question, then paste the first prompt above.

Once you’ve looked at the argument, evidence, and unanswered questions, request an Audio Overview for a specific listener. When it’s ready, compare one surprising claim and one important distinction with the original.

Familiar material is your advantage. You’ll notice what the audio captures, misses, simplifies, or overstates because you already know what should survive the trip.

Pick that document today. The goal isn’t to prove Google can make a podcast. It’s to find out whether listening helps you understand something well enough to use it.

Work from these documents.
Here’s what I need to decide or explain: [your question].
This is for: [your audience].
Explain the main argument and show the passages supporting its most important claims. Point out missing information, disagreements, and assumptions being treated as facts.
Separate what the documents say from your interpretation or advice. Don’t fill gaps with invented facts. If the evidence is weak, say so.
Before we make an audio overview, what should I check in the originals?

Create an engaging overview for business leaders who have not read The Comeback Code. Focus on the five comeback stages, explain why each stage matters, and use stories from the book when they clarify the lesson. Keep the tone practical rather than promotional. Do not introduce claims that are not supported by the selected sources.