Using AI to read long documents, contracts, and PDFs
Drop in a 40-page document and get the parts that matter — the decisions, the risks, the things you would otherwise miss. A practical workflow for using AI to read what you do not have time to read.
If you only ever use AI for one thing, let it be this: reading long documents that you do not have time to read yourself. Contracts, leases, terms of service, research reports, policy documents, long emails, transcripts, PDFs of any kind.
This is the use case where AI is most reliable, most underused, and most directly time-saving. Hallucinations are minimised because the model is working with text you provided. The synthesis is the actual work. And the difference between a thirty-page document and "the three things you need to know" is, on most workdays, one of the highest-leverage moves you can make.
This article is a three-prompt workflow that handles 90% of long-document tasks.
How to feed a document to AI
In every major AI assistant, there is a file-upload button — usually a paperclip or "+" near the input box. Drop in your PDF. For images of documents (scans, photographs), the model will OCR them automatically; for native PDFs, the text is extracted directly.
A few practical notes:
- PDFs over 100 pages may push against context limits on free tiers. Paid tiers (ChatGPT Plus, Claude Pro, Gemini Advanced) handle long documents better. For very long PDFs, try whichever tool on your plan is known for long context — at the moment that is usually Claude on paid tiers — and split the file if you still hit limits.
- Scanned PDFs work but lose precision on small or stylised text. For a contract you are about to sign, do not rely on AI OCR alone for the fine print.
- Multiple files at once are supported. You can upload three contracts and ask for a comparison.
- Email exports work — save a Gmail or Outlook thread as PDF and upload it.
The three-pass workflow
The mistake most people make is asking for "a summary" in one big prompt. You get a generic abstract that misses what you care about. Instead, do three short passes, each with a specific purpose.
Pass 1 — First impressions. What is this thing, and what are the three or four key takeaways?
Pass 2 — Risks and red flags. What in here could hurt me, surprise me, or cost me something?
Pass 3 — Decisions and actions. What do I actually need to do, decide, or ask about?
Do all three in the same conversation so the model can build on what came before.
Pass 1: First impressions
A reliable prompt:
I just uploaded a document. Before going deep, give me:
>
1. One sentence describing what this document is, who wrote it, and who it's for. 2. The three most important takeaways for someone reading it. 3. The structure — what are the main sections, in order, and what does each cover? 4. Anything that seems unusual or unexpected for this kind of document, given who I am.
>
About me: I am [your role, your situation, why you're reading this].
The "about me" line is doing real work. A contract review prompt from a freelance designer should produce different highlights than the same prompt from a corporate procurement officer. The model can calibrate when you tell it.
Read the response carefully. This pass usually answers the question "is it worth my time to read the whole thing?" Roughly 60% of the documents you would have plowed through can be safely skimmed once you have a sound pass 1.
Pass 2: Risks and red flags
This is the highest-value pass for contracts, legal documents, terms of service, employment agreements, leases, and anything where someone is asking you to commit. Continue in the same conversation:
Now read it again with one question in mind: what in here could hurt me or surprise me later?
>
List, in priority order:
>
1. Any clause that gives the other side a unilateral right — to terminate, change terms, change pricing, claim ownership, demand additional things from me, etc. 2. Any obligation on me that is open-ended or hard to comply with. 3. Anything that contradicts itself or is ambiguous in a way that would favour the drafter. 4. Anything common in this kind of document that is missing (a notable absence). 5. Numbers, dates, or terms that seem unusually high, low, or strict.
>
For each, quote the exact clause and explain in plain English why it matters. Mark anything you are unsure about with [unclear].
This is the prompt that justifies the model's existence. The model is unusually good at this kind of structured risk reading, and many knowledge workers — lawyers, accountants, HR, ops — treat the output as a first-pass highlight reel that saves them anything from twenty to ninety minutes per document.
A note: AI is not a substitute for legal advice when stakes are high. It is a brilliant first pass that catches the obvious things and helps you know what to ask a real lawyer about. That is exactly the right framing.
Pass 3: Decisions and actions
Now the operational pass.
Based on this document, give me:
>
1. The three decisions I need to make before responding or signing. 2. The two or three questions I should ask the other side to clarify before committing. 3. Any actions on me with specific deadlines or numerical thresholds (dates, amounts, response windows). 4. The single change to this document I would most want to negotiate if I could only ask for one.
>
Be concrete and specific. No "consider your options" or "consult a professional" platitudes — I want actionable next steps.
That last sentence is the difference between a useful answer and a CYA non-answer. Without it, models will hedge themselves into uselessness. With it, you get something you can act on.
After pass 3, you have a usable artefact: a one-page brief on a 40-page document, calibrated to your situation, with decisions, questions, and a negotiating angle.
Beyond contracts: other things this workflow handles
The three-pass structure generalises. Adjust the questions per pass but keep the shape.
Long research reports. Pass 1: takeaways. Pass 2: where is the data weakest, what are the unstated assumptions. Pass 3: what does this imply for my decisions.
Long email threads. Pass 1: who said what, who decided what, what is still open. Pass 2: where did anyone change their position, where is the disagreement hiding. Pass 3: who is waiting on what, who needs to do what next.
Annual reports. Pass 1: the three things that changed year over year. Pass 2: footnotes that contain the real information. Pass 3: what does this imply about the company's trajectory.
Long meeting transcripts. Pass 1: the decisions and the open questions. Pass 2: anyone who said something noteworthy that did not get picked up by the group. Pass 3: action items by person, with [unclear] tags for ambiguous owners.
Policy documents and regulations. Pass 1: who is covered, what is required, what is prohibited. Pass 2: exceptions, edge cases, conditions where the rules differ. Pass 3: what do I specifically need to do to comply.
In each case, the model is performing a kind of structured reading that humans are bad at when tired and pressured, which is when we do most of our reading.
Three things that lift quality further
Once the workflow is comfortable, three small additions push the output from useful to excellent.
Tell the model what kind of audience the summary is for. "Write the summary as if you were briefing a busy executive who has five minutes." Or "...briefing a careful lawyer who will check anything they're not sure about." Different framings produce different outputs.
Ask for the model's confidence. After each pass, "rate your confidence in each section from 1 to 5 and tell me which sections I should re-read myself to verify." This catches the parts where the model is hedging.
Use search mode for context. For a contract or a regulation, turn on web search and add to pass 2: "Where this document references external laws, standards, or precedents, briefly summarise what they say. Note anything that has changed in the last year." This gives you grounded context the model would not otherwise have.
The takeaway
A three-pass read of any document — first impressions, risks, decisions — takes about five minutes and is more useful than thirty minutes of skimming. It catches things you would miss, frames the document around your situation, and ends with actions you can take.
The next time you face a long document you have been putting off, do this once. The time you save and the things you catch will make it reflexive within a week.