How to ask questions to a PDF and get answers you can check

Updated 13 October 2026 · ChatPDF guides · 8 sources

AI can miss or misstate things. Summaries and answers from AI tools, ChatPDF included, can leave out important details or get them wrong. Check anything that matters against the document itself, especially legal, medical, financial and exam material.

Key points

If you know a word the answer contains, Find in the Preview app or in Notes is free, fast and exact. ^1,2
Chat-with-PDF tools search your document for a few relevant passages and let an AI model answer from them, so specific questions work best. ^3,4
Use the document’s own terms, ask one thing at a time, and ask what the document says rather than what is true in general.
Check answers you will rely on: AI models can state wrong things confidently. ⁵

"Asking a PDF a question" can mean two things: searching it yourself, or letting an AI tool read it and answer in plain language. Both are useful, and they work best together. Search is exact but only finds words you already know; AI handles questions phrased your way but can miss or misstate things. Here is how to get good answers from each.

First, try search (free)

Many questions are really "where does it say X?" On iPhone, open the PDF in the Preview app, tap the search button and type a word or phrase. Preview searches for the exact phrase, and you can turn on Whole Words or Match Case to cut down false matches ¹. If the PDF is attached to a note, the Notes app can search inside it too: open the note, choose Find in Note and select Include Attachments ². More tips are in how to search for text in a PDF on iPhone.

Search fails when you do not know the author's wording ("termination" versus "cancellation"), when the answer is spread across sections, or when you need an explanation rather than a location. That is where a chat-with-PDF tool helps.

How chat-with-PDF tools find an answer

Most of these tools use retrieval-augmented generation: the document's text is split into passages and indexed, the passages most similar to your question are retrieved, and a language model writes an answer from them ³. The index is built from numeric "embeddings" that place passages with similar meaning close together ⁶.

How a typical chat-with-PDF tool answers: the document is indexed once, then each question pulls a handful of matching passages for the AI model to read.

Two consequences follow. The model only sees the passages it is given, so a fact that sits in a passage the search did not pick can be missed. And a passage cut out of its context can be misread: a sentence such as "revenue grew 3% over the previous quarter" does not say which company or quarter on its own ⁴. Good questions make the search step easier. The details are in how chat-with-PDF tools work.

Question patterns that work

Instead of	Ask	Why it works better
"Tell me about the warranty"	"How long is the warranty, and what does it exclude?"	Specific, two clear parts to find
"Is this a good deal?"	"List every fee or charge the document mentions, with amounts."	Asks what the document says, not for a judgment
"What about cancelling?"	"What does the document say about cancellation or termination, and is there a notice period?"	Uses the likely terms and synonyms
"Explain everything in chapter 4"	"What are the three main arguments in the section titled 'Market risks'?"	Names the section by its own heading
"Is it true that X?"	"Does the document say X? Quote the sentence if it does, or say it is not mentioned."	Gives the model a way to say "not found" and you a sentence to search for

More templates, including follow-up chains, are in questions and prompts that work well for PDFs.

Tips that make a real difference

One question at a time. A list of five questions in one message tends to get shallow answers to each.
Use the document's vocabulary. If the policy says "member", ask about members, not customers.
Ask it to separate what the document says from general knowledge. "Answer only from the document" is a useful reminder, though the model can still stray from the text.
Follow up on vague answers. "Which section says that?" or "What exactly is the condition?" often surfaces the precise wording.
Be wary of answers about the middle of very long inputs. Studies of language models found they use information at the start and end of a long input more reliably than information in the middle ⁷.

Verify before you rely on it

AI models can produce fluent text that is not supported by their source, a problem researchers call hallucination ⁵. For anything that matters, take a key word, name or number from the answer, search for it in Preview, and read the full paragraph. If you cannot find support, treat the answer as unconfirmed. The step-by-step routine is in can you trust AI PDF summaries?

Where ChatPDF helps

ChatPDF is built for this kind of back-and-forth on iPhone and iPad: add a PDF from Files inside the app, then ask questions in plain language and follow up. Answers come in the language you ask in, even if the document is in another language, and you can copy them. The limits: it needs a PDF with selectable text, and page numbers are not shown in its answers, so locate the passage yourself before quoting it. See getting started with ChatPDF.

Common questions

Why did the AI say the document does not mention something that is in it?

Chat-with-PDF tools usually pass only a few retrieved passages to the model. If your wording does not match the passage, it may not be retrieved. Rephrase with the document’s own terms or name the section. ^3,4

Can I ask questions about a PDF in another language?

Many AI tools can answer in your language about a document written in another, which helps you understand it. For an exact translation of a passage, use a translation tool on the selected text instead. ⁸

Is search or AI more reliable?

Search is exact: it either finds the words or it does not. AI is more flexible but can be wrong. Use AI to find and explain, and search to confirm. ^1,5

Sources and further reading

Apple Support. Open and view PDFs and images in Preview on iPhone (iPhone User Guide). support.apple.com
Apple Support. Search through your notes on iPhone (iPhone User Guide). support.apple.com
Lewis P, Perez E, Piktus A, et al. (2020). Retrieval-augmented generation for knowledge-intensive NLP tasks. Advances in Neural Information Processing Systems 33 (arXiv:2005.11401). arxiv.org
Anthropic (2024). Introducing Contextual Retrieval. anthropic.com
Ji Z, Lee N, Frieske R, et al. (2023). Survey of hallucination in natural language generation. ACM Computing Surveys 55(12):248. doi:10.1145/3571730 (arXiv:2202.03629). arxiv.org
Cloudflare Developers. Vector databases (Vectorize reference). developers.cloudflare.com
Liu NF, Lin K, Hewitt J, et al. (2024). Lost in the middle: how language models use long contexts. Transactions of the Association for Computational Linguistics 12:157–173. doi:10.1162/tacl_a_00638 aclanthology.org
Apple Support. Translate text in apps on iPhone (iPhone User Guide). support.apple.com