Which AI models can read PDFs on Zeplik?

All of them, in one of two ways. Models from the Claude, GPT, Gemini, and Grok families receive the PDF as a real file and read the pages directly, including scans and figures. Every other model gets the complete extracted text, labeled page by page. You pick the model, and you can switch mid-conversation; the document stays available either way.

Can it read scanned PDFs?

Yes, on models that read PDFs natively: a scanned page is an image, and those models read it the way they read any image. On a text-only model there is nothing to extract from a scan, so Zeplik tells you plainly and suggests switching to a native-reading model instead of inventing content.

How long can the PDF be?

Native file reading covers documents up to 100 pages, which matches what the underlying models accept. Longer documents automatically fall back to full text extraction, so a 300-page manual is still readable and searchable in chat; you lose only the visual layer, not the words.

Does asking about a PDF cost extra?

Uploading and processing the document is free. You pay only the normal model-token cost of the conversation, exactly like any chat message, and the price is shown before you send. Follow-up questions about the same document are much cheaper than the first, because the document is cached rather than re-billed at full price each turn.

Is my document private?

Your file is stored privately in your account and sent to the model only to answer your questions. It is never shared with other users and never appears anywhere public. Delete the conversation and the attachment goes with it. For a session that leaves no trace at all, incognito mode runs the same chat without saving anything.

Can Zeplik create or edit PDFs too?

It creates them: ask for a report, a memo, or a summary as a PDF and Zeplik drafts a real, formatted PDF file you can download, with headings, tables, figures, footnotes, and a choice of layout templates. It does not modify an uploaded PDF's original bytes; instead it reads your document and produces a new file with the changes you asked for.

Chat with any PDF

Upload a PDF and ask about it like you would ask a colleague who just read it. Contracts, research papers, financial reports, manuals, scanned letters: Zeplik reads the actual document and answers from what it really says, with quotes and page references. On models that read PDFs natively, even scanned pages, charts, and tables are read directly from the page itself.

Try the summarizer

Uploading is free; you pay only the normal cost of the answer, shown before you send.

How it works

Upload a PDF and ask
Attach a PDF from the composer, or just drag it into the chat, and ask what you want to know: a summary, a specific answer, the key numbers, a clause. You can attach several documents in one message and ask across them.
Zeplik reads the real document
On models that read PDFs natively, the document goes to the model as an actual file, so scanned pages, charts, and layout are read directly. On every other model, Zeplik extracts the complete text page by page and labels each page, so answers can still point to where a fact lives.
Answer, quote, and follow up
The answer is grounded in what the document really says. Ask follow-ups in the same conversation without re-uploading; the document stays in context, and repeat questions are cheap because the document is cached rather than re-billed at full price.

Worked examples

Real prompts you can open in the composer with the text ready to review. Nothing sends on its own. Attach your own document and adapt the question.

Prompt: Summarize this PDF in one page, with the key numbers pulled out into a short table

The whole document is read, not just the first pages, so the summary reflects the conclusion and appendix as much as the introduction.

Open this in Zeplik

Prompt: What does this contract say about termination, notice periods, and penalties? Quote the exact clauses

Because the answer is grounded in the real text, you get the clause wording itself with its location, not a paraphrase from memory.

Open this in Zeplik

Prompt: Extract every table in this report and explain what each one shows

On a model that reads PDFs natively, tables and figures are read as they appear on the page, layout included.

Open this in Zeplik

Prompt: Compare these two papers: where do their findings agree, and where do they contradict each other?

Upload more than one PDF in the same message and ask across them; each stays separately labeled so quotes attribute to the right source.

Open this in Zeplik

What it can and cannot read

Two reading modes, one honest posture. Models that read PDFs natively see the page itself; everything else gets the complete extracted text, and Zeplik is plain about the difference.

Reads itReports, papers, contracts

Text PDFs read fully on every model. Zeplik extracts the complete text page by page, so answers can cite the page a fact came from.

Reads itScanned documents

On models that read PDFs natively (Claude, GPT, Gemini, and Grok families) scanned pages are read directly, like a person looking at the page. On other models a scan has no text to extract, and Zeplik says so instead of guessing.

Reads itCharts, figures, tables

Native-reading models see the actual page, so charts and complex table layouts are understood visually. Text-only extraction keeps table text but can lose the grid.

Reads itLong documents

Up to 100 pages travel to a native-reading model as a real file. Beyond that, or on other models, Zeplik falls back to page-numbered text extraction rather than dropping your document.

CannotPassword-protected PDFs

Encrypted files cannot be opened. Remove the password first, then upload.

CannotEditing the PDF in place

Zeplik reads PDFs and writes new ones (it can draft a fresh PDF or Word document from what it read), but it does not modify the original file's bytes.

Grounded, and honest about its limits

Answers come from the document. Quotes and page references trace back to the real text, and when something is not in the file, Zeplik says so rather than filling the gap from memory.
No silent degradation. If a scan cannot be read on your current model, or a very long document falls back to text extraction, the model is told exactly what it is looking at and tells you, including which model to switch to for the full picture.
Your document stays yours. Files are stored privately in your account, sent to the model only to answer you, and deleted with the conversation. Incognito mode keeps a session out of history entirely.

Frequently asked questions

Which AI models can read PDFs on Zeplik?: All of them, in one of two ways. Models from the Claude, GPT, Gemini, and Grok families receive the PDF as a real file and read the pages directly, including scans and figures. Every other model gets the complete extracted text, labeled page by page. You pick the model, and you can switch mid-conversation; the document stays available either way.
Can it read scanned PDFs?: Yes, on models that read PDFs natively: a scanned page is an image, and those models read it the way they read any image. On a text-only model there is nothing to extract from a scan, so Zeplik tells you plainly and suggests switching to a native-reading model instead of inventing content.
How long can the PDF be?: Native file reading covers documents up to 100 pages, which matches what the underlying models accept. Longer documents automatically fall back to full text extraction, so a 300-page manual is still readable and searchable in chat; you lose only the visual layer, not the words.
Does asking about a PDF cost extra?: Uploading and processing the document is free. You pay only the normal model-token cost of the conversation, exactly like any chat message, and the price is shown before you send. Follow-up questions about the same document are much cheaper than the first, because the document is cached rather than re-billed at full price each turn.
Is my document private?: Your file is stored privately in your account and sent to the model only to answer your questions. It is never shared with other users and never appears anywhere public. Delete the conversation and the attachment goes with it. For a session that leaves no trace at all, incognito mode runs the same chat without saving anything.
Can Zeplik create or edit PDFs too?: It creates them: ask for a report, a memo, or a summary as a PDF and Zeplik drafts a real, formatted PDF file you can download, with headings, tables, figures, footnotes, and a choice of layout templates. It does not modify an uploaded PDF's original bytes; instead it reads your document and produces a new file with the changes you asked for.

More on Zeplik

Upload a PDF, get the answer

Zeplik reads the real document and answers from it, quotes included. Free to start.

Chat with a PDF Browse all tools