PDF Key Points Extractor — 6-12 bullets with page citations and why citations matter
Sometimes you don't need a summary. You need the underlying claims, separated, each verifiable against the source. Key Points produces exactly that: a clean bulleted list of the document's substantive claims, each with a page citation so you can verify against the original. Here is how it differs from Summarize and TL;DR, why page citations are the load-bearing feature, and the three workflows where the bulleted-with-citations shape beats any prose summary.
What Key Points produces
Drop a PDF. The tool extracts the page text, runs a claim-extraction prompt, and returns a Markdown bullet list. Each bullet captures one substantive claim, finding, or argument from the document. Every bullet ends with a citation in [p. N] format pointing to the source page where the claim appears.
There are no intro paragraphs, no closing summary, no "in conclusion" prose. Just the claims, individually, scannable in 30 seconds. Total bullet count is typically 6-12 — shorter for brief documents, longer for substantial ones. The tool decides based on how many distinct claims actually exist; there is no fixed cap.
Why page citations are the load-bearing feature
The citation suffix on every bullet is what separates Key Points from a generic bullet-summary tool. Three concrete uses:
- Verifying the AI's interpretation. Every bullet is checkable. If a claim seems surprising, jump to that page in the source and verify the AI extracted it correctly. The citation makes verification a one-click operation rather than a search.
- Pulling source-attributed quotes for downstream use. When you cite the document in your own writing, you need to point to a specific page. Key Points already tells you which page each claim came from.
- Catching hallucinated claims. If a bullet has no citation (or cites a page that doesn't contain the claim), that's a flag for a fabricated finding. The page-citation discipline forces the model to ground every claim — and when it can't, that's information too.
When bullets-with-citations beats prose
Three workflows where the bulleted shape is the right one:
- Building a research summary. You're compiling findings from multiple documents into a literature review or report. Key Points from each document gives you a citation-ready bullet list to weave into your own synthesis. Prose summaries would require re-extracting individual claims; bullets are already in the right format.
- Note-taking for self-study. Reading a textbook chapter or research paper for class. Bullet points are easier to revisit than prose summaries. Page citations let you jump back to the source for context when the bullet alone isn't enough.
- Briefing someone else. Need to brief a colleague on what's in a document. Bullets are scannable; they can read the list in 30 seconds and decide what to ask about. Prose forces them to read sequentially.
How the tool decides what's a 'key point'
Three criteria the model uses:
- Substantive claims, not topical mentions. "The paper discusses inflation" is not a key point. "Inflation rose 3.2% in Q4 2024 [p. 14]" is. Specifics over generalities.
- Self-contained, not context-dependent. Each bullet should be readable on its own without requiring the reader to remember the previous bullet. Cross-referential bullets get collapsed or split.
- Decision-relevant, not decoration. A bullet that captures a finding affecting the reader's understanding or decisions gets kept. Throat-clearing observations get filtered out.
When Key Points isn't the right shape
Three cases where a different summary tool fits better:
- You want narrative flow. If you need the document's argument to flow as a story — context → claim → evidence → conclusion — bullets break the flow. Use Summarize / Standard for connected prose.
- Short triage is enough. If you only need to decide whether to read the document at all, use TL;DR (2 credits vs Key Points' 3). The bullet structure is overkill for the triage decision.
- Studying for an exam. Exam prep benefits from spaced-repetition flashcards more than from a bullet list. Use PDF to Flashcards for that workflow.
Reading the output critically
Three habits when working with Key Points output:
- Verify high-stakes bullets at the source. If a bullet captures a claim you'll cite or rely on, click through to the cited page and confirm the bullet captures the source accurately. AI extraction is high-quality but not infallible.
- Watch for missing citations. Bullets without [p. N] are flags — the model couldn't tie the claim to a single page. Sometimes that's legitimate (multi-page themes), sometimes it's hallucination. Worth verifying explicitly.
- Use the bullet list as a navigation aid. Beyond reading the bullets, use them as a reverse index. "I remember a claim about X — here's the bullet — page 47 is where to look in the source." Faster than scanning the PDF directly.
Limits and pricing
PDF Key Points charges 3 credits per document. The tool handles PDFs up to 25 MB. Processing runs on our servers; the document is in memory only during extraction and is never persisted. Output is a Markdown bullet list with per-bullet page citations.
Common pairings: Key Points for the bullet-list view; Summarize / Standard for the prose view; PDF Chat for follow-up questions on specific bullets. The three together cover most reading workflows.