Google's NotebookLM Now Builds PowerPoint Decks for You

Google's NotebookLM has added presentation generation capabilities, and the pitch is seductive: feed it some sources, type what you want, and get a complete slide deck without touching PowerPoint. The update rolled out this week, adding a "Revise" button that lets users edit presentations through text prompts rather than manual formatting.

Julian Goldie, an SEO consultant who runs YouTube tutorials on AI tools, demonstrated the features in a walkthrough that shows both the promise and the peculiar selling style that now accompanies every AI product launch.

The mechanics are straightforward. Users upload documents, links, or other sources to NotebookLM. The system generates a presentation based on that material. If you want changes—different colors, cleaner layouts, reorganized sections—you type instructions rather than dragging elements around a canvas. "You don't need any design skills. You don't need any technical skills. There's no worrying about PowerPoint anymore," Goldie explains. The system processes the request and regenerates the deck.

NotebookLM offers two research modes: "fast" pulls from five to ten sources quickly, while "deep" casts a wider net. Users can choose between presenter slides with key talking points or comprehensive decks with full text. The tool exports to PDF, which matters for anyone who needs to actually share these presentations outside Google's ecosystem.

What Sets It Apart

The distinction between NotebookLM and Google's own Gemini chatbot illustrates how AI tools are fragmenting into specialized purposes. Goldie addresses this directly: "When you're using something like Notebook LM, it's really like a media platform, right? So, you can generate videos, you can generate slideshows, you can generate audio and podcasts and you can't really do that as well inside something like Gemini."

NotebookLM generates infographics, data tables, mind maps, audio overviews, and now slide decks from the same source material. It's positioned as an organizational hub where multiple formats derive from a single research base. Gemini, by contrast, remains conversational—better for coding assistance or quick answers than structured content production.

The audio feature deserves mention because it's genuinely unusual. NotebookLM can create podcast-style conversations about your sources, complete with two AI voices discussing the material. It's either fascinating or deeply weird, depending on your tolerance for synthetic personalities explaining your research back to you.

The Productivity Calculation

Goldie frames the update in terms of time savings: "The old way was like you would spend hours researching the topic manually, opening up Canva or PowerPoint and start from scratch, manually typing out every single word, hiring a designer to make it look good, exporting it, realizing the formatting is broken, and then trying again, still ending up with something that looks quite generic and forgettable, and the whole process would take like 4 to 8 hours minimum."

That's the standard AI pitch—automation converts hours to minutes. But the calculation assumes the bottleneck is mechanical execution rather than thinking. If your problem is "I need slides but lack design skills," tools like this solve it. If your problem is "I'm not sure what story this data tells," the automation might just produce formatted confusion faster.

The demo shows Goldie generating a presentation about OpenClaw (an AI tool) while simultaneously creating other content types. The system works in parallel—you can queue up a slide deck, an infographic, and a data table, then return when they're done. For certain workflows, particularly in training materials or client reports where format matters more than narrative innovation, that parallelism has value.

The Design Question

How good do these automated presentations actually look? Goldie's examples appear competent—clean layouts, reasonable color choices, standard corporate aesthetics. They won't win design awards, but they also won't embarrass you in a team meeting. "The quality comes from your thinking and inputs. The tool handles the design. That's the upgrade," he argues.

This positions design as a solved problem, which designers might dispute. The tool applies templates and formatting rules. It doesn't understand visual hierarchy in the way a human with taste does. But for the majority of business presentations—sales decks, training modules, project updates—template-level competence suffices.

The question isn't whether AI-generated slides beat hand-crafted presentations by skilled designers. It's whether they beat what most people actually produce: text-heavy slides assembled at midnight before a morning meeting.

The Catch

Goldie's demonstration doubles as a lengthy advertisement for his "AI Profit Boardroom" community, which offers courses, coaching calls, and daily updates on AI tools. This isn't unusual—most AI tutorial content now comes wrapped in course offers—but it does mean evaluating the tool through a sales filter.

The update is free, which matters. Google's strategy with NotebookLM appears focused on adoption rather than monetization at this stage. How long that lasts depends on usage patterns and whether Google sees a path to revenue that doesn't involve selling access or training data.

The tool also inherits NotebookLM's fundamental limitation: it's only as good as its sources. Garbage in, formatted garbage out. If you feed it shallow research, you get professional-looking slides that say nothing interesting. The automation accelerates production; it doesn't inject insight.

Where This Goes

Presentation software has changed remarkably little since PowerPoint launched in 1987. We've added animations, templates, and cloud collaboration, but the basic model—manual slide construction—persists. Tools like NotebookLM represent a genuine shift: presentations as outputs generated from source material rather than artifacts built from scratch.

Whether that shift improves communication depends on what presentations are actually for. If they're documentation—records of decisions, training materials, project summaries—automation helps. If they're persuasion—sales pitches, conference talks, investor presentations—the human element matters more. A tool that generates slides can't yet understand why a particular metaphor lands with a specific audience, or when to pause for effect.

Goldie's enthusiasm is characteristic of AI adoption right now: every update gets positioned as transformative, every tool as indispensable. But the update is real, the functionality works, and for people who need to produce presentation materials regularly, it's worth examining beyond the sales pitch.

The question isn't whether AI can generate slides. Clearly it can. The question is whether removing friction from slide creation leads to better communication or just more slides.

Bob Reynolds has covered technology for five decades and has seen seventeen different "PowerPoint killers" come and go.