Why most NotebookLM prompts fail
You’ve uploaded your sources. You ask: “Summarize this.”
NotebookLM gives you a summary. It’s fine. But it’s not what you actually needed — which was the specific argument in chapter 3, or the data point that contradicts the author’s conclusion, or a list of every claim that isn’t backed by a citation.
The difference between a mediocre and a great NotebookLM session is prompt specificity. These 17 prompts are designed to extract real research value.
Stage 1: Understanding your sources
Use these when you’ve just uploaded documents and need a map of what you’re working with.
1. The contradiction finder
“What claims in these sources directly contradict each other? List each contradiction with quotes from both sides.”
2. The evidence auditor
“List every major claim in these sources. For each, tell me: is it backed by data, by citation, or by the author’s opinion only?”
3. The gap finder
“What questions do these sources raise that they don’t answer? What’s conspicuously absent?”
4. The assumption surfacer
“What assumptions does this source make that it never explicitly states? What would need to be true for its conclusions to hold?”
5. The methodology critique
“If this were submitted for peer review, what methodological objections would a skeptical reviewer raise?”
Stage 2: Synthesizing across sources
Use these when you have multiple sources and need to find connections.
6. The consensus map
“Across all my sources, what do all authors agree on? What does only one author claim? Format as a table.”
7. The timeline builder
“Extract all dates and events mentioned across my sources. Build a chronological timeline.”
8. The stakeholder mapper
“Who are the key stakeholders mentioned across these sources? What does each want, fear, and stand to gain or lose?”
9. The definition reconciler
“The term ‘[X]’ appears in multiple sources. How does each source define it differently? Which definition is most precise?”
10. The surprising connection
“What’s the most surprising connection you can draw between sources that appear unrelated on the surface?”
Stage 3: Preparing outputs
Use these when you’re ready to turn your research into something — a report, article, presentation, or decision.
11. The executive brief
“Summarize the 5 most important things a time-pressed executive needs to know from these sources. No more than 3 sentences per point.”
12. The steelman builder
“Give me the strongest possible version of the argument [X]. Use only evidence from my sources. Make it as convincing as possible.”
13. The devil’s advocate
“Now argue the opposite position. What’s the strongest case against [X]? Use evidence from my sources.”
14. The FAQ generator
“Generate 10 questions someone unfamiliar with this topic might ask. Answer each question using only my sources.”
15. The decision framework
“I need to decide whether to [decision]. Extract all relevant evidence from my sources and organize it as: evidence for, evidence against, unknowns.”
Stage 4: Quality checking
Use these before finalizing any research output.
16. The citation checker
“For every factual claim I’ve made in [pasted text], tell me: does my source material support it, contradict it, or neither?”
17. The recency auditor
“Which facts and figures in my sources are most likely to be outdated? What would I need to verify before publishing?”
Saving prompts with Kortex
Retyping these every session is tedious. Kortex’s Prompt Library lets you save any prompt, organize them by category, and insert them with one click — directly inside NotebookLM. For a full walkthrough of Kortex’s features, see the getting started guide.