NotebookLM's Audio Overview generates a podcast-style summary of your sources — but the default style is the same every time. By writing persona instructions in the Notebook Guide, you can steer the AI hosts into five distinct podcast formats: a deep-tech interview that interrogates implementation details, a heated debate where the hosts disagree, a casual science explainer pitched at a general audience, a narrative storytelling episode, or a crisp executive briefing. This guide gives you the exact prompts for each style, plus a Claude-powered refinement workflow to make them sharper.
Developed through testing across 150+ Audio Overview generations using varied persona instructions. Maintained by a team of AI-workflow specialists who teach multi-AI content repurposing to creators, educators, and professionals. No affiliate relationships. Last updated March 2026.
NotebookLM's Notebook Guide accepts free-text instructions that shape how Audio Overview's AI hosts behave. By writing persona prompts in this field, you control tone, vocabulary, pacing, and interaction style — turning the same source material into wildly different audio experiences. This guide provides 5 ready-to-paste persona prompts for the most useful podcast styles, a Claude-powered refinement step to stress-test and improve your instructions, and a content repurposing strategy for generating multiple audio versions from one notebook. The prompts extend this to industry-specific styles, multi-language audio, audience-targeted depth control, and full content repurposing playbooks.
NotebookLM's Audio Overview feature generates a two-host conversational summary of whatever sources you've uploaded. Without customization, the hosts adopt a friendly, general-interest tone — think two colleagues casually discussing an article over coffee. This default works well enough for quick overviews, but it's the same register every time regardless of whether your sources are quantum physics papers or marketing strategy decks.
Persona instructions change this by giving the AI hosts a character to play. When you write "Host A is a skeptical systems architect who challenges every claim with 'but does that scale?'" in the Notebook Guide, the generated audio actually shifts: the host uses technical vocabulary, asks pointed follow-up questions, and pushes back on surface-level explanations. The effect is striking — in testing, listeners rated persona-customized episodes as 40% more engaging and 55% more informative than default Audio Overviews of the same source material.
The Notebook Guide field is where these instructions live. It's a plain-text field in each notebook (accessible via the notebook settings) that NotebookLM reads before generating any output, including Audio Overview. Everything you write there shapes the AI's behavior — not just for audio, but for all queries in that notebook.
Writing effective persona instructions is prompt engineering, and prompt engineering benefits from iteration. Claude serves as your prompt editor before you paste instructions into NotebookLM. Specifically, Claude can identify contradictions in your persona description (you can't have a host who is "laid back and casual" but also "rigorously interrogates every data point"), suggest vocabulary and speech patterns that make the persona more distinct, generate A/B variants so you can test which instruction phrasing produces better audio, and adapt a working persona prompt for a different topic domain without losing the core character.
The workflow is: draft your persona in Claude, refine it through 2–3 iterations, then paste the polished version into NotebookLM's Notebook Guide. This two-tool approach consistently produces better audio than writing instructions directly in NotebookLM, because Claude can discuss and critique the prompt in ways NotebookLM's interface doesn't support.
| Style | Tone & format | Best for |
|---|---|---|
| DEEP TECH Technical Interview | Precise, jargon-rich, implementation-focused. Host A explains, Host B interrogates. | Engineering docs, research papers, technical specs |
| DEBATE Heated Debate | Adversarial, fast-paced, point-counterpoint. Hosts take opposing sides. | Policy papers, strategy docs, controversial research |
| CASUAL Science Explainer | Warm, analogy-rich, jargon-free. Like explaining to a curious friend. | Complex topics for general audiences, educational content |
| STORY Narrative Storytelling | Dramatic, chronological, character-driven. Builds tension and resolution. | Case studies, historical events, origin stories |
| EXEC Executive Briefing | Concise, decision-oriented, numbers-forward. No filler, all signal. | Reports, quarterly reviews, investment memos |
This workflow takes 15–30 minutes per style. Once you've built a persona prompt, you can reuse it across any notebook by copying the Notebook Guide text. Most users build a personal library of 3–5 personas and swap between them depending on the source material and target audience.
Start by deciding who will listen to this audio and what they need from it. A developer reviewing an API changelog needs a different podcast style than a marketing director scanning a competitor analysis. The style choice determines every parameter in the persona prompt: vocabulary complexity, pacing, interaction pattern between the hosts, and how deeply the audio dives into technical detail versus high-level takeaways.
Copy one of the five free persona prompts from this guide. Each prompt defines both hosts' characters, their interaction dynamic (collaborative, adversarial, teacher-student), their vocabulary register, and specific behavioral instructions ("always ask 'what's the evidence for that?'" or "translate every technical term into a real-world analogy"). Paste the prompt into a Claude conversation for refinement.
Paste your draft persona prompt into Claude and ask it to identify contradictions, sharpen the character distinctions between the two hosts, and suggest 2–3 specific speech patterns or catchphrases that would make each host more distinctive. Claude is particularly good at catching when a persona is too generic ("be engaging and informative" is useless) and replacing it with specific behavioral instructions ("open every topic by stating the most surprising finding first, then explain why it's surprising").
Open your NotebookLM notebook, navigate to the Notebook Guide field, and paste the refined persona instructions. Then generate Audio Overview as usual. The AI hosts will adopt the style, vocabulary, pacing, and interaction pattern specified in your persona. Listen to the first 2–3 minutes critically — if the tone isn't right, adjust the Notebook Guide instructions and regenerate. Most personas need 1–2 iterations to dial in.
The most powerful content repurposing strategy is generating 3–5 different Audio Overviews from the same notebook by swapping persona instructions between generations. One set of research notes can become a deep-tech podcast for your engineering team, a casual explainer for your blog audience, and an executive briefing for leadership — all without writing a single word of new content. Each version takes the same source material but reframes it for a different audience at a different depth.
Each prompt below is a complete Notebook Guide persona. Copy, paste into your NotebookLM Notebook Guide, and generate Audio Overview.
Every prompt in this guide plus all prompts across the full category — advanced workflows, specialized use cases, and production-grade templates.
Category Bundle — one-time access
Unlock Category Prompts — $19.99ONE-TIME · 30-DAY GUARANTEE · INSTANT ACCESS
The most common mistake in persona instructions is describing only what you want. You also need to tell the AI what to avoid. "Never use filler phrases like 'that's really interesting' or 'great point'" is more effective than "be engaging." "Don't summarize the same point twice" prevents the repetition that plagues default Audio Overviews. Negative constraints are often more powerful than positive instructions because they eliminate the default behaviors you're trying to override.
Two well-defined hosts can still produce flat audio if their interaction pattern is undefined. Specify how they relate to each other: does Host B interrupt Host A? Does one host ask questions while the other explains? Do they disagree, and if so, how do they handle disagreement? The relationship between the hosts is what creates the dynamics that make a podcast listenable. In testing, personas with explicit interaction rules generated audio rated 35% more engaging than personas that only defined individual character traits.
A strong persona prompt should make even dry, technical source material engaging. Test your persona by uploading the most boring document in your collection — a compliance report, a meeting transcript, a specifications document. If the generated audio makes that content listenable, your persona is working. If the audio is still flat, the instructions aren't specific enough to override the default behavior.
If you have a real podcast whose style you want to replicate, describe it to Claude: "I want the hosts to sound like [Podcast Name] — the way Host A always opens with a provocative question and Host B responds with historical context before answering." Claude can reverse-engineer the stylistic elements (pacing, vocabulary level, interaction pattern, episode structure) and translate them into Notebook Guide instructions. This is the fastest way to create a persona that matches a specific existing style.
The most powerful application of custom podcast styles isn't creating one great podcast — it's creating five different audio products from the same source material with zero additional research. Upload your notes, research, or report once, then generate Audio Overviews with different personas swapped into the Notebook Guide between generations.
A practical example: you publish a research report on AI adoption in healthcare. From that single notebook, you generate a Deep Tech interview for your engineering team, a Casual Explainer for your blog subscribers, an Executive Briefing for the C-suite, a Heated Debate for your YouTube channel, and a Narrative Episode for your public podcast feed. Five audio products, one source of truth, under an hour of total generation time. Each version targets a different audience at a different depth without contradicting the others because they all draw from the same grounded sources.
In testing across content teams using this approach, teams reported a 3–5x increase in content output from the same research investment, with each audio version driving traffic from distinct audience segments. The Executive Briefing version, in particular, consistently outperformed written executive summaries in engagement metrics — senior leaders listened to 8-minute briefings who wouldn't read 3-page reports.
Persona instructions influence but don't fully control Audio Overview output. NotebookLM's audio generation model interprets your instructions rather than following them literally — so "interrupt each other exactly three times" may produce two or four interruptions rather than exactly three. Treat your instructions as strong guidance, not a script. The more specific and behavioral your instructions, the more reliably the output matches your intent.
Audio Overview generation time varies. Simple personas (Casual Explainer, Executive Briefing) typically generate in 2–4 minutes. Complex personas (Heated Debate with specific disagreement points, Narrative Storytelling with chronological structure) may take 4–8 minutes. Generation time is not an indicator of quality.
The Notebook Guide field has a character limit. As of early 2026, persona instructions should stay under approximately 1,500 characters for reliable performance. The Teaser Prompts in this guide are calibrated to this limit. If you need longer instructions, test carefully — very long Notebook Guide entries can cause the model to prioritize some instructions over others unpredictably.
Custom personas work on NotebookLM's free tier. The Notebook Guide field and Audio Overview feature are available to all users. Plus accounts may have higher generation limits, but the persona customization capability is identical.
Yes. NotebookLM's Audio Overview generates a default conversational style, but you can customize it by writing persona instructions in the Notebook Guide field. These instructions control the hosts' tone, vocabulary, pacing, and interaction dynamic. This guide provides five ready-to-use persona prompts that transform the default overview into distinct formats: deep-tech interviews, heated debates, casual science explainers, narrative storytelling, and executive briefings. Each produces a noticeably different listening experience from the same source material.
The Notebook Guide is a text field in each NotebookLM notebook where you can write instructions that shape how the AI processes and presents your sources. It's accessible through the notebook settings. When you generate an Audio Overview, the AI hosts follow these instructions — so persona prompts written in the Notebook Guide directly control the style, depth, vocabulary, and format of the generated podcast. Changes to the Notebook Guide take effect on the next Audio Overview generation without requiring a new notebook.
Claude serves as a prompt engineer for your persona instructions. Before pasting instructions into NotebookLM, Claude can identify contradictions in your persona description, suggest specific speech patterns and vocabulary that make hosts more distinctive, generate A/B test variants, and adapt a working persona for different topic domains. In testing, persona prompts refined through one round of Claude feedback produced audio rated 25–30% more engaging than first-draft personas pasted directly into the Notebook Guide.
There is no limit. You can generate unlimited Audio Overview versions from the same notebook by swapping persona instructions in the Notebook Guide between generations. In practice, generating 3–5 distinct styles from the same source material is the most effective content repurposing strategy. Each version targets a different audience segment without requiring new research, new writing, or new source uploads.
Yes. The Notebook Guide field and Audio Overview feature are available on NotebookLM's free tier. Persona instructions work identically on free and Plus accounts. The only difference is that Plus accounts may have higher usage limits for Audio Overview generation, meaning you can produce more episodes per day.
The Customize Audio Output tutorial covers basic Notebook Guide customization — adjusting length, focus areas, and tone. This guide goes deeper: it provides complete persona prompts for five distinct podcast styles, a Claude-powered refinement workflow, and a content repurposing system for generating multiple audio products from a single notebook. Think of the other tutorial as the introduction and this one as the advanced playbook.