AI: understand documents
The idea: most documents are too long to read closely and too important to ignore. These five AI apps let you absorb a document — ask it questions, condense it, turn it into actions — while keeping a path back to the original page. They change nothing in your content (except Abstracts, which adds one page); they produce new material that is about your document.
All five read the text layer of your PDFs. That makes them powerful on digital documents and blind on scans, until you run OCR — with one exception: Abstracts recognises scanned text itself.
The five apps
| App | Takes | Gives | One-liner |
|---|---|---|---|
| Ask PDF | 1–10 PDFs + a question | Cited answers, follow-ups | Chat with documents; every claim links to a page |
| PDF Knowledge Base | PDFs (add more later) | A linked wiki (.zip of Markdown) + cited chat | A collection you build up and question again and again |
| Abstracts | PDF(s) | The PDF with an Abstract page added | Five styles, three lengths, any language |
| Cheat Sheet | PDF(s) or just a topic | One dense printable PDF | Definitions, formulas, facts, dates, summary |
| Checklist | PDF(s) + a goal | Prioritised list with source pages | Tasks, learning, preparation or analysis |
Choosing quickly
- I have a question → Ask PDF
- I have many documents and will ask again over weeks → PDF Knowledge Base
- I want a summary I can keep or send → Abstracts
- I want something to print and pin up → Cheat Sheet
- I want to know what to do → Checklist
- My PDF is a scan → Searchable first (except for Abstracts)
Rules of thumb
- Text first. If you cannot select the text, the AI cannot read it. See good inputs.
- Say what you want. A specific question or topic beats a generic one.
- Verify what matters. Click the citation, re-check the number. See trust and verification.
- Long documents are sampled. Free runs have a time limit; read the progress messages and FAQs.
- Download what you need. Uploaded files and knowledge bases are deleted automatically after a short time.
Continue with writing good inputs, trust and verification, or try the lab. For ready-made chains see Understand a long document fast.