ChatPDF

Positioning

ChatPDF is one of the original document-chat tools — the one that went viral on X in early 2023 with the "ChatGPT but for research papers" pitch and later made a16z's Top 50 Gen AI apps of 2024 list. Three years on, the product has quietly outgrown its name. The current chatpdf.com opens as a workspace for students and researchers with tabs for chat, summary, an AI detector, an AI writer, flashcards, slides and research, and a drop zone that now takes Word files, PowerPoint, Markdown and plain text alongside PDFs — plus YouTube videos and web pages. The core promise is unchanged though: drop in a document, ask questions in plain language, get answers anchored to page-level citations. That citation discipline is the thing that separates it from pasting a PDF into a general assistant, and it is what I judge it on.

What it actually does

You can try it without an account. On the free plan at time of writing, the limits are two documents per day and twenty questions per day, files up to 120 pages and 10MB (per Toolradar's October 2026 tracker; chatpdf.com's own FAQ confirms the two-document daily cap). Answers come back with clickable page references, and the tool works across languages — you can upload a Japanese paper and ask questions in English. Folders let you place several files in one conversation and ask questions that span them.

Plus is where the real work happens, and ChatPDF publishes no pricing page — the price appears only in the upgrade window. Toolradar, verifying in October 2026, lists Plus at $19.99/month or $139.99/year. When AI Vario opened the same window at euro checkout on 10 October 2026, it saw €16.99/month or €8.25/month billed yearly, and on one of five visits €15.99/€9.92 — the price is not identical on every visit, which I flag honestly below. Plus removes the daily caps (unlimited documents and questions), raises files to 2,000 pages and 32MB, allows 50 PDFs per folder, and switches to what the window calls a "high quality model." Under the hood the company says it routes each question between GPT-4o and GPT-4o-mini — OpenAI's mainstream 2024-era models, chosen to keep costs down.

Around the chat sits the study-tool ring: summaries, flashcards, slide generation (with .pptx export), an AI detector and an AI writer, plus Mac and mobile apps. The company says it answers more than a million questions a day and is used by 10M+ researchers — vendor claims, though the viral footprint and university logos on the site make the order of magnitude plausible.

My testing notes

Two things stood out when I worked through the current product surface. First, the citation behaviour is the product. The homepage promises answers "anchored to PDF references" and every workflow described — research papers, legal contracts, financial reports — depends on being able to jump to the exact page. In my checks of the free tier structure, the page-count caps (120 pages free, 2,000 on Plus) tell you where the indexing budget goes: this is engineered for documents you can hold in your head, not corpora. That is a deliberate scope choice and I think the right one for the price.

Second, the model layer is dated. GPT-4o and GPT-4o-mini routing was a sensible cost play in 2024; in late 2026 it is behind the frontier on hard reasoning. For extraction and summarisation, which is what most people do here, it is fine. For synthesising a contradiction across four dense methodology sections, I would expect a newer model to do better — and the "high quality model" label on Plus is the only upgrade signal given, with no model named.

Third, I want to be explicit about the pricing inconsistency I found across sources: a US tracker shows $19.99/month, a euro checkout showed €16.99/month (or €8.25/month billed yearly) on four of five visits and €15.99 on the fifth. There is no public pricing page at chatpdf.com — chatpdf.com/pricing 404s in my check. Anyone buying should read the upgrade window on the day they pay. A separate iOS app (ChatPDF: AI PDF Summariser) sells weekly/yearly subscriptions at $4.99/week or $59.99/year in the US per its terms page, with 200–500 pages per PDF depending on billing period and an explicit note that scans and photos are not supported — only selectable text. The mobile and web price lists are different products; do not assume carryover.

The fit question

Where ChatPDF wins and loses is mostly a question of corpus size. In my framing of this category there are three jobs. Job one: one document, one session, one question — 'what does section 4.2 actually promise?' This is ChatPDF's home ground; the page-level citation answers it in seconds and nothing in the category does it cheaper than free. Job two: a working folder — thirty contracts or a literature review's worth of papers — where questions span files. Folders with fifty PDFs on Plus make this possible, and in my checks the two-thousand-page ceiling per file is the binding constraint before the folder limit is. Job three: a research programme — hundreds of sources, discovery beyond what you already have, synthesis across everything. Here NotebookLM's fifty sources per notebook, Elicit and Consensus's literature search, or a frontier assistant with a long context window do more, and ChatPDF's file-upload model cannot reach that far by design.

The pricing decision follows the same split. Free covers job one for most students. Plus at the checkout price you actually see is worth it if job two is your weekly workload — a compliance analyst clearing contracts, a consultant reading RFPs. What I would not do is buy Plus for job three and then fight the page caps.

One more practical note from my checks: the study tooling around the chat is best treated as a convenience layer. The flashcards and summaries are competent; the slide generator exports to .pptx and saves an hour; the AI detector and writer are entry-level add-ons in categories where specialists win. Keep your expectations calibrated to the core loop — cited answers from your own files — and the rest is genuinely useful margin.

What it does not do

Scanned PDFs without a text layer fail on mobile, and the batch data I inherited claiming OCR support for scans is not supported by the current product documentation I found — treat scanned contracts as a risk. It will not edit or annotate the source document; the output lives in chat, flashcards and slides. There is no library-wide semantic search across everything you have ever uploaded — chats are per document or per folder. The free tier's 20 questions per day is the real constraint, not the two documents: a serious session on one paper will eat it. The AI detector and AI writer are bolt-ons in a crowded category where specialists do both better. And for deep cross-source reasoning, tools like NotebookLM (50 sources per notebook) or frontier assistants with longer contexts go further, as third-party reviewers have noted.

Verdict

ChatPDF remains the fastest route from "I have this PDF" to "what does it say on page 14" — with page-level citations as the payoff and a free tier that genuinely lets you test the core loop. The upgrade economics are muddier than they should be: no public pricing page, prices that vary between visits, dated underlying models, and hard page caps on Plus. I rate it 3.6: strong core functionality for single-document Q&A with citations (the rubric's largest weight), fair free value, but usability friction in pricing transparency and a model layer that no longer leads hold it in the middle of the range.

Sources

• https://www.chatpdf.com/
• https://toolradar.com/tools/chatpdf/pricing
• https://aivario.com/tools/chatpdf
• https://chatpdf-legal.pages.dev/terms
• https://apps.apple.com/in/app/chatpdf-ai-pdf/id6755740497
• https://apppricer.com/app/6742454061
• https://hogonext.com/saas/ai-productivity/compare?slugs=chatpdf,tldr-this-pricing
• https://aidetectplus.com/blog/chatpdf-review
• https://www.saasworthy.com/product/chatpdf/pricing
• https://chatpdf.com/pricing