
ChatPDF
ChatPDF lets you upload a PDF, Office file, YouTube video or web page and question it in plain language, with every answer anchored to the page it came from. Two documents daily are free, no account needed.
What is ChatPDF?
ChatPDF is a document-reading assistant: you hand it a file and then simply ask it questions. It arrived in March 2023 as one of the first mainstream tools of its kind, and its central idea has not changed since - the conversation and the document sit side by side, and every answer carries clickable references such as [P12] that jump straight to the page the claim came from. That anchoring is the whole argument: it turns a plausible-sounding summary into something you can check in a second.
The input is broader than the name suggests. Beyond PDF, it accepts .doc, .docx, .ppt, .pptx, .md and .txt, runs OCR over scanned pages, and will just as happily take a YouTube video, a web page or text pasted straight from the clipboard. Files can be grouped into folders and queried together, so you can compare several papers or contracts in one thread. Language is handled loosely on purpose: upload a document in one language and ask about it in another, with an interface available in thirty-nine locales. Under the hood, queries are routed dynamically between GPT-4o and GPT-4o-mini depending on what each one needs.
Around that core sits a small suite. PDF Summary produces structured summaries that follow the document's own shape. AI Research searches the literature, ranks journals by Scimago quartile, and builds cited summaries from abstracts, claiming better than 90% accuracy while limiting full-paper chat to open-access work. AI Writer offers contextual autocomplete, generates whole sections, keeps citations synchronised and exports to Word or Markdown. An AI Detector scores text for machine authorship. Flashcards and slides are generated from any document.
There is also a public REST API at api.chatpdf.com/v1, with endpoints to add a source by URL or upload, send chat messages, stream replies and delete sources; files are capped at 2,000 pages or 32 MB. Distribution covers the web, a desktop application, iOS and Android. The company claims more than 10 million users and over a million questions a day, and describes SSL in transit, encryption at rest and a SOC 2 Type II certified storage provider.
What it does
- Question a PDF, Word document, slide deck or text file in plain language and get answers with clickable page references
- Summarise a whole document, a single chapter or a report while keeping its original structure
- Put several files in one folder and interrogate them together, comparing and cross-referencing them in a single conversation
- Search academic literature, filter by year and by journal quartile, then read cited summaries
- Draft essays and papers with source-backed autocomplete and citations that stay in sync
- Test whether a piece of text was written by AI, with a confidence score across more than twenty languages
- Turn a document into flashcards or slides, or chat with a YouTube video or a web page
When to use ChatPDF / When not to
A quick filter to help you decide if ChatPDF is the right fit.
When to use ChatPDF
- Researchers working through scientific papers, journal articles and books, who need findings they can trace back to a source
- Students revising for exams, working through homework and turning course material into flashcards or slides
- Teachers, markers and editors who need to judge whether a submitted text was written by a machine
- Professionals navigating contracts, financial reports, manuals and training material without reading them cover to cover
- Developers who want document question-answering inside their own product, through a documented REST API with an explicit commercial licence
When not to use ChatPDF
- Organisations bound by HIPAA, FISMA or GLBA, which the terms of service exclude by name
- Procurement teams that require a signed data processing agreement and a published list of subprocessors, neither of which exists
- Anyone who must keep documents inside the European Union, since the terms place hosting in the United States
- Students hoping to outsource graded work, which the terms treat as grounds for reporting to their institution
- Heavy daily readers unwilling to subscribe, because the free tier stops at two documents a day
How to use ChatPDF
A typical end-to-end flow, from setup to results.
- Open chatpdf.com - no account is required to begin
- Drop in a file, browse for one, or paste a link or raw text with Ctrl+V
- Choose your source: PDF, Word, PowerPoint, Markdown, plain text, a YouTube video or a web page
- Wait for indexing, then type your question into the chat panel
- Click any [P] reference to jump to the cited page in the document pane beside the conversation
- Press Summarize in the chat for a structured overview that mirrors the document's own layout
- Create a folder, add several files to it, and question them all in one conversation
- Switch languages with the picker, or simply ask ChatPDF to answer in the language you prefer
- Create a free account to keep your history and sync it across devices
- For programmatic use, obtain a sourceId from sources/add-url or sources/add-file, then call chats/message
Pros & Cons
Pros
- Usable in seconds, with no account and no card - the free tier is permanent rather than a countdown
- Page-anchored citations make verification immediate, which is the honest answer to hallucination
- Input coverage well beyond PDF, taking in Office files, YouTube videos and web pages
- Genuinely multilingual end to end: the document and the conversation need not share a language
- A whole reading workflow in one place - summarise, search, draft, detect, revise
- Available on the web, desktop, iOS and Android, with progress carried between them
- A documented API with an explicit commercial licence, unusual at this price point
Cons
- No pricing anywhere on the site: /pricing returns a 404 and offers are served dynamically after sign-in, so the only published figures come from the mobile stores
- The free tier is narrow at two documents a day, which a single afternoon of reading exhausts
- All purchases are non-refundable, and cancellation only takes effect at the end of the paid term
- Hosting is placed in the United States even though the publisher is German, and the terms obtain consent for that transfer
- No DPA, no subprocessor list, and the SOC 2 Type II certification belongs to an unnamed storage provider rather than to ChatPDF
- No documented way to exclude your documents from model training, and no explicit statement that they are not used for it
- The legal paperwork contradicts itself: minimum age 13 in the terms against 18 in the privacy notice, two different addresses and two different VAT numbers
Pricing & Plans
ChatPDF operates a permanent free tier rather than a trial: two documents may be analysed each day, with no registration required, and the research module is currently offered without a usage limit. A single paid tier, ChatPDF Plus, removes the document ceiling and unlocks the advanced features. No price is published anywhere on the website - the pricing path returns a 404 and offers are served dynamically once a user signs in. The only published figures are those on the publisher's own mobile store listings, where the monthly Plus subscription is priced at USD 14.99 and the annual plan at USD 89.99. Payment is handled by Stripe and accepts Visa, Mastercard and American Express. A separate API plan exists, but its price is not disclosed. Purchases are non-refundable.
- two documents analysed per day
- no registration needed
- a free account adds saved history and multi-document conversations
- USD 14.99 per month
- unlimited document analysis and advanced features (price from the official App Store listing)
- USD 89.99 per year (price from the official App Store listing)
- a separate subscription for API access
- no price published
Data, GDPR & hosting
A consolidated view of how ChatPDF handles your data.
GDPR overview
ChatPDF is published by a German company, so the GDPR applies to it by law, not by choice, and the privacy notice engages with it directly: it names the GDPR and the UK GDPR as applicable, sets out the legal bases relied on - consent, legal obligation, vital interests - and lists the rights available in the EEA, the UK and Canada, from access and rectification to erasure, restriction, portability and objection. Requests go through a hosted form or a dedicated privacy@chatpdf.com address, and readers are pointed to the European supervisory authorities and the Swiss regulator. What is missing matters too: no data processing agreement, no subprocessor list, no standard contractual clauses, no data protection officer, and Do-Not-Track signals are explicitly ignored. No Article 27 representative is named, which is correct for an EU-established company. The notice itself carries an effective date of 5 January 2023.
Who owns the data?
The terms are unusually clear on ownership: ChatPDF GmbH asserts no ownership over what you upload, and you keep full title and any intellectual property rights attached to it. What the company keeps is its own side of the service - the software, design, content and marks. Two caveats temper the reassurance. Uploaded material may be treated as neither confidential nor proprietary, and the company reserves the right to edit, recategorise, screen or delete anything you submit, at any time and without notice. The privacy notice adds that your information may change hands in a merger, asset sale, financing or acquisition. Finally, the terms disclaim all liability for data loss or corruption, despite routine backups, so the only copy you can rely on is your own.
Reuse rights
For the web product, the licence is deliberately narrow: non-exclusive, non-transferable and revocable, granted for personal, non-commercial use or an internal business purpose. You may download or print portions of the content for that purpose, but any wider commercial exploitation needs written permission first, requested by email. Systematically harvesting content to build a database, compilation or directory is forbidden without written consent. The important exception runs the other way: the API carries an explicit commercial licence, letting you fold document question-answering into your own operations and paid offerings, provided you stay within the same terms. Nothing in the terms restricts what you do with the answers the tool generates for you. The research module adds one editorial rule of its own - cite the original paper, never the tool.
Data retention & training
Hosting summary
The clearest statement comes from the terms of service, which say plainly that the services are hosted in the United States and obtain the user's consent to transfer and process data there. That sits somewhat awkwardly with the rest of the arrangement: the publisher is established in Berlin, German law governs the terms, and disputes go to the European Court of Arbitration in Strasbourg. On the security side, the company describes SSL in transit and encryption at rest, and refers to a storage provider holding SOC 2 Type II certification - the provider is not named, and the certification is theirs rather than ChatPDF's. Payments are handled by Stripe, which stores all payment data. There is no published subprocessor list, no data processing agreement and no mention of standard contractual clauses. One technical caveat worth stating: the domain resolves to an anycast address belonging to its hosting platform's content delivery network, which geolocates to a United States city. That is an edge node, not a storage location, and should not be read as a data centre.
Things to keep in mind
Risks and trade-offs to weigh before adopting ChatPDF.
- Reading the summary instead of the source is a habit that forms quickly, and it quietly erodes the deep reading and critical judgement the tool is supposed to support
- Answers are still model output: the research module itself claims better than 90% accuracy, which means it is wrong often enough to matter, and its summaries are built from abstracts rather than full texts
- The AI detector reports a false positive rate of one in ten thousand, but an accusation of machine authorship built on a score alone remains contestable and can do real harm to a student or an author
- The terms forbid academic dishonesty and reserve the right to report violations to the institution concerned, so the shortcut carries a genuine consequence
- Uploading a confidential or third-party document puts the obligation on you, and the terms place hosting in the United States with no opt-out from model training documented
- Non-refundable purchases combined with prices that appear nowhere on the site make for a financial commitment you cannot examine before signing in
- Look-alike apps crowd both mobile stores under the same name, so an unwary download can send private documents to an unrelated publisher
Setup & Integrations
Technical difficulty
Effectively none for normal use. There is nothing to install and nothing to configure: open the site in a browser, drop in a file, ask a question. No account is needed for the core features, and a free one only adds saved history. Desktop, iOS and Android applications install the usual way. The only path that asks anything technical is the REST API, and even there the work is small - obtain a key, add a source by URL or upload, then post chat messages to one endpoint. Call it none for the web product, low for the API.
Deployment
Apps stores
Integrations
Supported languages
Behind ChatPDF
Social
Resources
All the official URLs gathered for verification and reference.
Alternatives
Tools that compete with or complement ChatPDF.
Frequently asked questions
Do I need an account to use ChatPDF?
Which file types does it accept?
Is ChatPDF free, and what does the paid plan cost?
Which AI models does it use?
Can I question several documents at the same time?
Are my documents used to train AI models?
Where is my data hosted?
Is there an API?
What is the minimum age?
Who publishes ChatPDF, and can I get a refund?
Should you pick ChatPDF?
ChatPDF was among the first tools to make a document answer questions, and it has spent the years since turning that trick into a workflow. The feature that still matters most is the least flashy one: answers carry references to the pages they came from, and a click takes you there. For anyone whose work depends on being able to defend a claim - a researcher, a student writing up, an analyst reading a report - that single design choice does more than any amount of fluency. Around it sits a genuinely useful suite: structured summaries, literature search ranked by journal quartile, source-backed drafting, AI text detection, flashcards and slides. It runs on the web, on the desktop, on both mobile platforms, and behind a documented API that is explicitly licensed for commercial use.
The reservations are real and they cluster in one place: everything commercial and legal is underdocumented. No price appears on the site at all; the figures quoted here come from the publisher's own app store listings, and the web product serves offers dynamically, so what you are shown may differ. Purchases are non-refundable. Hosting is placed in the United States despite a Berlin publisher, there is no data processing agreement and no subprocessor list, and nothing states whether uploaded documents feed model training. The paperwork itself disagrees with itself on the minimum age, the company address and even the VAT number, and the privacy notice has not been revised since January 2023.
The verdict splits cleanly by use. For an individual reader - a student, a researcher, a professional working through their own documents - this is a strong, mature and inexpensive tool, and the free tier is enough to judge it. For a regulated organisation, the gaps need closing before anything sensitive is uploaded.
- Choosing a selection results in a full page refresh.
- Opens in a new window.