Google’s NotebookLM: The Game-Changing Tool for Fiction Writers

Consider a science fiction novelist toiling on a complex series, juggling dozens of characters and story threads. After finishing a chapter, he might upload it into Google’s new AI-powered tool called NotebookLM and prompt a chat with the AI about what’s working and what isn’t. “I have them discuss what I’ve written and offer a reader’s perspective on it,” the writer explains.⁴ It’s akin to having an impartial beta reader on call at any hour. In an online forum he even confessed that he’s reached a point where he “can’t imagine [his] workflow without these tools.”⁵

Such enthusiasm is no longer an outlier. In a recent survey of over 1,200 authors by BookBub, nearly 45% of authors reported using some form of generative AI in their process⁶, whether for researching details, brainstorming, marketing copy or, increasingly, editing and proofreading their drafts. Half still swore off AI as if it were witchcraft (some invoking a near-Butlerian Jihad against “thinking machines,” to borrow Frank Herbert’s term for a holy war on AI).⁷ But behind closed doors many writers are quietly experimenting, discovering that an algorithmic assistant can sharpen their prose and catch inconsistencies long before a human editor ever lays eyes on the manuscript.

A Beta Reader That Never Sleeps

Rewriting a rough draft is often called the place where the real writing happens: a painstaking process of refining style, tightening structure and fixing story issues. Traditionally, writers might rely on beta readers or advance readers (often patient friends or fellow writers) to flag what’s confusing or dull in a second draft. Now, AI is stepping into that role as an ever-alert, always-available early reader. Unlike a human friend, the AI beta reader never tires or sugarcoats its feedback, and can be summoned at 2 a.m. to dissect a scene that isn’t landing right.

How does an AI offer feedback? One approach is simply asking a general chatbot (like OpenAI’s ChatGPT or Anthropic’s Claude) to read a chapter and comment on the tone, pacing, or clarity. Many authors have done this and found the results both eye-opening and sobering. “I ask ChatGPT for feedback on my chapters or outlines, on tone, voice, pacing, or specific scene details,” says novelist Camilla Monk, describing her editing process.⁸ The AI will happily enumerate which plot points shine and which fall flat. It might point out, for instance, that a secondary character seems underdeveloped, or that a chase sequence dragged in the middle.

In Monk’s experience, ChatGPT acts like “the sum of all average readers,” offering a kind of market-average perspective on the story.⁹ That can be extremely valuable: a machine-mediated glimpse into how a typical reader might perceive the narrative. If the AI finds a sequence incoherent, chances are real readers would too.

But there’s a flipside. By definition, an AI’s sensibility is generic, trained on billions of words of published text. As Monk notes, if she were to “strictly implement all of its suggestions, I will never achieve more than an average work of commercial fiction.”¹⁰ In other words, an AI reader’s instincts tend toward the conventional. It will never have the eccentric genius of an Ezra Pound, who pushed Eliot into daring territory; instead it will nudge you toward the safest, most broadly acceptable choices. That makes it a great diagnostic tool, a way to catch glaring issues, but a questionable arbiter of style. “Don’t let it think for you. Let it be the assistant,” Monk advises, “but ultimately… YOU do the thinking.”¹¹

Still, as a kind of supercharged beta reader, AI proves remarkably useful. Authors are using it to simulate sensitivity readers (e.g. “Does my portrayal of a 13-year-old sound authentic?”), to gauge whether a plot twist was too obvious, or simply to summarize their own story back to them to see if the themes are coming through. Advanced models can even be prompted to role-play as a harsh critic or a supportive fan, depending on what flavor of feedback the writer needs. In the old days, a writer might coax two trusted friends into giving very different perspectives on a draft. Now you can replicate that at will: ask the AI to critique your ending like a snarky New York Times reviewer, then in the next query have it gush like a die-hard fan explaining what they loved.

One of the more surreal but effective new techniques is having AI literally talk about your book in audio, as if you were listening to a podcast review. Google’s NotebookLM, for example, can generate an audio overview where two voices have a natural-sounding conversation about the document you uploaded.¹² Writers have used this to hear their story discussed as though by book critics or even by characters from the story itself. Robb Wallace, an indie fantasy author, recounts using NotebookLM to produce a debate between “two characters [from his novel] reviewing the book’s pros and cons from a good-cop, bad-cop perspective,” an experience he found “very refreshing and a great way to make improvements.”¹³

Inside NotebookLM: A Second-Draft Power Tool

While general AI chatbots can critique and brainstorm, Google’s NotebookLM deserves special attention in this new landscape of literary tech. Introduced in 2023 as “Project Tailwind” and now an experimental product in Google Labs, NotebookLM is essentially a personal AI research assistant that you can train on your own notes and documents.¹⁴ Think of it as a bespoke chatbot that has read only what you feed it, whether that’s your novel draft, your series bible of characters and lore, or a folder of research PDFs and world-building notes. This source-grounding means its knowledge and advice stay tethered to your material, drastically reducing the risk of off-base tangents or fabricated (“hallucinated”) facts.¹⁵

For an author embarking on a second draft, especially of a sprawling, complex story, NotebookLM can be a game-changer. Robb Wallace describes his writing process for an epic fantasy series as a “controlled chaotic frenzy” of scattered notes: Google Docs chapters, character bios, historical timelines, even hand-scribbled maps.¹⁶ Keeping track of a vast imaginary world had become a nightmare of manually cross-referencing details. “I have been spending endless hours re-reading, looking for that specific world-building detail, lore, [or] character conversation,” he admits.¹⁷

Upon discovering NotebookLM, he effectively dumped the entire 150,000-word first book of his series, the 85k-word draft of book two, and 25k words of miscellaneous notes into the AI. The tool indexed everything into a centralized, interactive knowledge base for his saga.¹⁸ Suddenly, he could query his own fiction the way one might query a database. With a quick question, NotebookLM could recall a minor character’s eye color or unearth where in book one a prophetic verse was mentioned, details that would take a human hours to skim for, but which the AI retrieved in seconds.¹⁹

More impressively, Wallace used NotebookLM to spot plot gaps and dropped threads across his series. He asked it to analyze book one and his work-in-progress sequel for any logical inconsistencies or unresolved elements.²⁰ Sure enough, the AI identified two major aspects of an initial prophecy that hadn’t been followed up in book two, threads the author had been “slowly getting to,” but which a reader might reasonably expect to see addressed sooner.²¹ “At 85k words into book two, I realized that the reader may want this info quicker,” Wallace writes. He promptly went back to weave those elements into earlier chapters of the sequel, adding scenes to address the “very obvious gaps” the AI had flagged.²²

The appeal of NotebookLM goes beyond just error-catching. It aims to be an all-purpose creative research hub. Each “notebook” can accept up to 50 documents (Google Docs, PDFs, web pages, even YouTube transcripts) and turn them into a conversational knowledge base.²³ The free tier generously offers 100 such notebooks, each with up to half a million words of content, which is ample for even the most verbose novelist.²⁴ Once your draft and notes are uploaded, you can:

• Generate summaries of long chapters or entire manuscripts, complete with key topics and suggested questions to explore.²⁵ This can help a writer see the forest for the trees, useful if a storyline has grown convoluted.
• Ask detailed questions about your text. For example: “List all the scenes where my protagonist mentions his childhood,” or “What are the key clues leading up to the big twist?” The AI will scan the sources and respond with answers drawn only from your story, often quoting lines with citations.²⁶ It’s like having a superfan reader with photographic recall of everything you’ve written, ready to be quizzed.
• Create outlines, timelines and mind maps. In NotebookLM’s interface, you can prompt it to produce a timeline of events from your novel or a mind-map of character relationships.²⁷ If your plot has multiple flashbacks and parallel threads, the timeline view can reveal if the sequence of events actually makes sense. The mind-map might show how every character is linked, potentially exposing a lonely subplot that never connects to the main story.
• Generate custom reports. Want a thematic analysis of your book? Or a report on each character’s arc? The AI can be instructed to compile that, again citing your text. Authors have used this to ensure that, for instance, every supporting character has a complete arc or that the themes they intend are consistently reinforced throughout the narrative.
• Offer style and grammar suggestions. Because NotebookLM (and similar tools) are built on large language models, they inherently know a thing or two about syntax. You can ask it to highlight clunky sentences or passive voice in your prose, much like a supercharged Grammarly. The difference is that NotebookLM’s suggestions are grounded in context, it knows the broader story, so it might even say, “This paragraph describing the weather is unusually florid compared to the rest of your tone,” a nuanced stylistic catch that simple grammar software wouldn’t manage.

Perhaps the most intriguing use cases are those that mimic high-level editorial feedback. NotebookLM allows you to ask it to adopt specific personas or “hats” when giving a response.²⁸ You could say: “Read my short story and critique it as if you are a Pulitzer-winning literary critic,” or conversely, “Summarize this chapter in the voice of a snarky teen reader on Goodreads.” Because it’s grounded in your actual text, the criticism will point to real parts of your story. This persona feature is powerful for customizing the tone of feedback. Wallace points out you can have it speak as a “grammar nazi” focusing on technical errors, or as a “developmental editor” focusing on big-picture structure.²⁹ By switching hats, a writer can essentially triangulate the truth of their draft’s issues.

Mapping the Manuscript and Managing the Chaos

Fiction writers often create vast imaginative worlds with elaborate backstories, timelines, and rules, and then face the herculean task of keeping all that straight through revision after revision. The more sprawling the project, the more likely the second draft devolves into a hunt for errors: Did I already use this symbolism before? Does the heroine’s eye color accidentally change between chapters? When was the last time we saw this minor character? These continuity questions can drive a lone writer mad. AI offers a kind of outsourced memory: a way to map the entire manuscript and all its ancillary data in one place.

For instance, an author of historical fiction might upload her research notes and first draft into NotebookLM and ask, “Generate a timeline of all events in my novel alongside real historical events of that year.” The AI can align the plot’s timeline with factual history, potentially exposing anachronisms or implausible gaps. A fantasy author might request, “Show me a list of all magical spells mentioned, and which characters use them,” quickly revealing if any Chekhov’s gun (a magical object introduced early) was forgotten later. This kind of macro-level analysis is something human editors certainly do, but it’s laborious. Here it happens with a quick query. As one tech reviewer marveled, after uploading sources you can “literally talk to your entire research library at once,”³⁰ treating the AI like an omniscient lorekeeper for your own work.

Even beyond text, NotebookLM supports multimedia inputs. You can throw in images (like a hand-drawn map of your fictional world) or even the link to a private YouTube video of your book trailer, and the AI will incorporate those into its knowledge base.³¹ This hints at a future where an AI could cross-reference your written story with visual materials, for example, ensuring a town’s map layout matches the directional cues in your prose. Meanwhile, more straightforward benefits abound. By having perfect recall of every detail you’ve provided, the AI spares you the painstaking task of flipping through pages or skimming countless files to find one reference. And because it is source-grounded, if it answers a question, you can click to see the exact excerpt in your draft or notes where that answer came from,³² double-checking it in context.

For writers whose projects have spiraled into complexity, this kind of AI assistance can feel like emerging from a cave of chaos into clear daylight. “Goodbye, digital chaos!” as one author gleefully wrote after integrating NotebookLM into his workflow.³³ Instead of drowning in loose papers and lost sticky notes, writers are regaining control over their creations, able to navigate and query their story worlds with unprecedented ease. This doesn’t just save time; it often surfaces new creative connections. Ask the AI to, say, list themes or motifs across all your notebooks, and it might reveal that your seemingly unrelated subplots each echo a pattern (perhaps themes of isolation or resilience) that you hadn’t consciously stitched together. Recognizing that can inspire you to emphasize it more in the next draft, turning disjointed pieces into a more coherent tapestry.

The AI Editor’s Custom Fit: Tailoring Technology to Your Voice

As seduced as one might be by these shiny new tools, wise authors approach them with caution and craftsmanship. Perhaps the biggest fear among literary writers is that AI tools could homogenize their voice, sanding off the quirky edges that make their style distinctive. There’s some justification for that worry. Feed a few pages of a brash, idiosyncratic narrative into a generic AI and ask for a rewrite, and you might get back something grammatically polished but soul-sucking, all the flavor boiled off. Early adopters have learned to avoid that trap by being very intentional in how they use AI. “I specifically tell it: I don’t want you to rewrite, I just want to know how you perceive it,” notes Camilla Monk, who has learned to corral ChatGPT when it overeagerly starts suggesting actual line edits.³⁴ By customizing her prompts, she gets feedback without letting the AI hijack her prose.

Customizing an AI editor often means setting the stage in the prompt. If you simply paste a chapter and say “Edit this,” the AI might default to a bland style guide mode. But if you say, for example: “You are a celebrated writing coach. Here is my chapter; give me feedback on pacing and emotional impact, without changing the text,” you are more likely to get a thoughtful critique that respects your voice. With NotebookLM, such customization can also involve selecting which of your source documents the AI should draw on for a given question.³⁵ For instance, if you want thematic feedback, you might include your outline and a note you wrote on the story’s themes as sources, so the AI focuses on those, rather than, say, detailed historical research notes that aren’t relevant to theme.

Another form of customization is training the AI on examples of feedback you like. Although NotebookLM itself is closed-box (it doesn’t let you fine-tune the underlying model on new data), some writers have taken an extra step with other AI systems: they compile a document of their favorite writing advice or past editorial letters they’ve received, and include that in the prompt context. The AI then tries to emulate that style of critique. For instance, you might show it an excerpt of a famous Stephen King essay on writing, and then ask it to critique your horror story in a similar blunt, witty manner. This hack can produce eerily apt feedback at times, as if you solicited comments from King or Ursula K. Le Guin.

Maintaining one’s literary voice is ultimately a manual effort, though. Editors like Tiffany Yates Martin caution that at this stage “AI… works against [freeing your voice], homogenizing the author’s voice” if allowed to take the reins.³⁶ The consensus among practitioners is to use AI in the analysis phase, not the actual rewriting. Have it highlight issues, generate alternatives, even draft a quick example of how one might rewrite a clunky paragraph, but then incorporate those insights in your own way, in your own words. One author likens it to having a “devil’s advocate” on your shoulder: useful for pointing out weaknesses, but not someone you’d let actually write your book for you.³⁷

Even in areas like proofreading, where one might assume the AI could safely autocorrect grammar and move on, authors have learned to double-check. Large language models can sometimes be confidently wrong, misapplying a grammar rule or “fixing” a nonstandard dialect in dialogue that was intentionally written that way. Therefore, some writers use AI-driven grammar checkers (like the latest Grammarly or Microsoft Word’s new AI “Copilot”) as a first pass to catch obvious typos, but then carefully review each change. The AI never sleeps or loses focus, but it also doesn’t know your intent; that’s why a human brain must stay in the loop.

Augmentation, Not Abdication

Every innovation in writing technology, from the typewriter to the spell-checker, has faced a period of hand-wringing about the death of craftsmanship. AI is no exception. There are heated debates in writing communities and publishing circles about the ethics and aesthetics of machine assistance. Yet, in practice, many authors are finding that using AI for the second draft and beyond feels less like cheating and more like having a turbocharged set of reference manuals plus a brutally honest friend who’s read absolutely everything you’ve written. It’s not so much writing by committee as writing with a very odd, very well-read imaginary friend.

Crucially, the writer remains the mastermind. The AI might suggest twenty ways to rephrase a sentence, but only the author can decide which (if any) fits the character’s voice and the story’s mood. The AI might flag a plot point as illogical, but it’s the author who devises a clever solution that feels true to the narrative. In this sense, what AI offers is a form of intellectual companionship in the lonely process of revision. It’s there to bounce ideas off of, to argue with (“No, you dumb bot, I want that scene to be confusing at first, it’s intentional!” one might find oneself saying), and to prod you when you get complacent.

For writers on a budget or without access to professional editors, AI tools can also democratize quality control. Not everyone can afford a seasoned developmental editor or multiple rounds of beta readers. But a free (or modestly priced) AI service can provide a decent simulation of those early editorial passes, potentially raising the baseline quality of self-published or first-time authors’ manuscripts. “I am considering using ChatGPT to edit the manuscript I worked hard to write, because the quotes for human editing are so high,” one writer confessed, sparking discussion on social media.³⁸ Many others likely quietly do just that. Of course, the best scenario is to use AI and still hire a human editor, arriving with a cleaner draft means the human can focus on higher-level improvements.

Will the future of fiction be co-written by algorithms? Probably not in the literal sense. The heart of storytelling remains an innately human art, rooted in lived experience and imagination. But it’s increasingly likely that behind many polished novels, there will have been an AI or two in the room, helping the author untangle the knots of their early drafts. Already, nearly half of authors admit to using AI in some capacity,³⁹ even if few trumpet it from the rooftops. One day, using an AI assistant in writing might be as unremarkable as using spell-check or Google Docs’ autocomplete.

As for that specter of homogenization, it’s a valid concern that the literary world must guard against. But one could argue that artists have always absorbed and reacted to the prevalent technologies of their time. Just as the arrival of word processors led to longer, more complex manuscripts (since editing became easier), the arrival of AI might lead to fiction that is more internally consistent, richly detailed, and perhaps more experimental (since authors can validate wild ideas or alternate story paths quickly with the AI before committing). The individual voice will remain paramount; the best writers will use the tools to amplify, not erase, what makes their style singular.

In a fitting twist of fate, if Eliot and Pound’s collaboration happened today, it might have been mediated by a neural network. We’ll never know how that hypothetical AI-edited poem would compare to the masterpiece that emerged in 1922. But in our current reality, writers large and small are beginning to find that their second drafts shine a little brighter with a silicon second reader by their side, an algorithmic editor, armed with endless patience and reams of insight, helping human creativity reach its fullest potential.

Broken English stuff:

What’s all this chatter about “romantasy”? The Allure of Supernatural Academies in modern young adult literature How to write a novel in the age of AI Must writers now collaborate with AI to stay ahead? The Truth About Writer’s Block: Why Discipline Beats Inspiration Every Time Write Dangerously and Authentically: Blueprints of Storytelling: Why Outlines and Treatments Matter