Treat PDFs as Documents, Not Pasted Fragments
Use file inputs and page-aware prompts for PDFs, scans, and tables.
For documents, page-grounded extraction should usually come before explanation. The move Document understanding is a page-grounding problem as much as a text problem. Headers, tables, footnotes, page references, and nearby clauses often resolve ambiguity. When you paste fragments, you strip away those cues and make the answer harder to audit. Why it works The answer quality improves because the model receives the right evidence in the right role. Instead of improvising over an incomplete text prompt, it can inspect, compare, extract, or transcribe against a defined standard. Common trap Do not add modalities for style. Add them only when they…
Sign up free — one personalized lesson every day, matched to your role and goals.
Already have an account? Sign in