Use Claude's Deep Research to explore any topic with exhaustive depth, then pipe the synthesized findings through NotebookLM to produce a logically structured, visually coherent presentation deck — complete with narrative arc, speaker notes, and imagery direction. From curiosity to keynote in under an hour.
Research is divergent — you want to explore widely, follow threads, and surface surprising connections. Presentation is convergent — you need a single narrative thread, a logical sequence, and a visual rhythm that holds attention. Most people try to do both simultaneously, and neither comes out well. The research is shallow because you're already thinking about slide layouts, and the deck is scattered because you haven't finished thinking.
Claude's Deep Research solves the first half. It conducts multi-step, autonomous web research — following chains of sources, cross-referencing claims, and producing a comprehensive synthesis report with citations. You get the depth of a research assistant who spends hours in a library, delivered in minutes.
NotebookLM solves the second half. Upload the Deep Research report as a source, and NotebookLM treats it as grounded truth — it can restructure, summarize, and extract presentation-ready content without hallucinating beyond what the research actually found. The RAG architecture ensures every slide traces back to evidence.
Combined, they produce something neither tool creates alone: a presentation deck that is both deeply researched and compellingly structured, with every claim grounded in verified sources.
Frame the topic as a specific question or angle — not just "climate change" but "how are Pacific Island nations adapting infrastructure to sea level rise?" Specificity determines research depth.
Submit the question to Deep Research. Claude autonomously searches, reads, cross-references, and synthesizes — producing a structured report with citations, key findings, and evidence quality notes.
Copy or export the Deep Research report into a new NotebookLM notebook. Generate a Briefing Doc and Mind Map to establish the conceptual landscape. This becomes your grounded source of truth.
Ask NotebookLM to extract the narrative arc: what's the story this research tells? What sequence of revelations leads to the core insight?
Ask NotebookLM to extract the 8–12 strongest data points, statistics, and quotable findings — with source citations for each.
Bring the narrative arc + evidence inventory back to Claude. Ask it to design a slide-by-slide architecture: title, key message, supporting evidence, visual direction, and speaker notes per slide.
Upload Claude's slide architecture back into NotebookLM. For each slide, ask NotebookLM to produce the exact text, pulling only from the grounded research. This prevents any hallucination from entering your final deck.
Return to Claude with the completed slide content. Ask for image generation prompts per slide — whimsical, magical, data-viz, or photographic styles. Claude also polishes speaker notes and transitions for presentation flow.
Build the deck in your presentation tool (Google Slides, PowerPoint, Keynote). Run a final check in NotebookLM: "Does this deck accurately represent the source research? Flag any slide that overstates or misrepresents a finding."
Research question: "What are the most distinctive New Year's traditions around the world, and what cultural values do they reveal?"
Deep Research output: 4,200-word synthesis covering 23 countries, with citations from cultural anthropology sources, travel journalism, and UNESCO cultural heritage documentation.
Final deck: 16 slides — opening with a global map of midnight timestamps, then grouped by theme (fire traditions, water traditions, food traditions, noise traditions), each with whimsical illustrated imagery and grounded cultural context.
| Phase | Tool | Why this tool |
|---|---|---|
| Deep exploration | Claude Deep Research | Autonomous multi-step web research with source chaining |
| Source grounding | NotebookLM | RAG ensures answers only come from uploaded research |
| Narrative extraction | NotebookLM | Briefing Doc + Mind Map reveal the story structure |
| Slide architecture | Claude | Structural reasoning, logical sequencing, 200K context |
| Slide copywriting | NotebookLM | Grounded text generation — no hallucination risk |
| Imagery & polish | Claude | Creative direction, image prompts, speaker notes |
| Final verification | NotebookLM | Source-level fact-check against original research |
The quality of your deck is determined by the quality of your research question. A vague question ("tell me about AI") produces a vague deck. A precise question ("how are K-12 school districts in the US integrating generative AI tools into math instruction, and what early outcomes are emerging?") produces a deck with authority. Spend 5 minutes getting the question right — it's the highest-leverage moment in the workflow.
Submit your question to Claude's Deep Research feature. It will autonomously search the web, read full articles, follow citation chains, and produce a comprehensive report. Read it carefully — this is your only chance to add supplementary questions or redirect the research before it becomes your deck's foundation. Look for gaps, surprising findings, and the strongest data points.
Create a new notebook and upload the Deep Research report (copy-paste or export as PDF). If you have supplementary materials — your own notes, related papers, prior presentations — add those too. Generate NotebookLM's automatic Briefing Doc and Mind Map. These pre-built summaries reveal the conceptual structure of your research before you start designing slides.
Run two targeted queries in NotebookLM. First: "What is the narrative arc of this research? What story does the evidence tell — what's the setup, the tension, and the resolution?" Second: "List the 10 most compelling data points, statistics, and quotable findings in this research, with source citations for each." These two outputs become the raw material for your slide design.
Bring the narrative arc and evidence inventory to Claude (ideally in a Project with the research uploaded as context). Ask Claude to design a slide-by-slide architecture: slide number, title, key message (one sentence), supporting evidence (specific data point or finding), visual suggestion, and speaker notes. Claude's structural reasoning makes it excellent at sequencing arguments for maximum impact.
Upload Claude's slide architecture back into your NotebookLM notebook. For each slide, prompt NotebookLM to write the exact on-slide text using only information from the research sources. This is the critical grounding step — NotebookLM can't hallucinate because it only draws from your uploaded material. Ask it to include a source citation for each slide's key claim.
Return to Claude with the completed slide content. Ask for image generation prompts that match your desired visual style — whimsical watercolor illustrations, clean data visualizations, dramatic photography direction, or magical/fantastical imagery. Claude also polishes speaker notes, adds smooth transitions between slides, and flags any slides that feel weak or redundant.
Copy any prompt below. Replace bracketed placeholders with your own details.
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
Deep Research is available on Claude Pro ($20/month) and produces one report at a time — expect 3–5 minutes for complex queries. NotebookLM is free, with higher limits available through Google AI Plus ($19.99/month). If you're producing decks regularly, both subscriptions pay for themselves after 2–3 projects compared to manual research time.
The handoff between tools involves copy-paste — there's no direct integration yet. Keep a consistent naming convention for your exports: "[Topic]_deep_research_[date].md" and "[Topic]_slide_architecture_[date].md" so you can trace the provenance of any claim in your final deck.
This workflow is optimized for evidence-driven presentations — keynotes, briefings, lectures, pitches, and reports where credibility matters. For creative presentations (mood boards, brainstorms, artistic showcases), you may want to lean more on the imagery direction step and less on the grounding verification.
Image generation prompts from Step 7 can be used with any image generator — DALL·E, Midjourney, Ideogram, or Claude's own image generation. The prompts are tool-agnostic. For whimsical or magical visual styles, Midjourney and Ideogram tend to produce the most distinctive results; for clean infographics, DALL·E often works best.