PDF to JSON
Extract PDF content into structured JSON format.
Drag & drop your PDF here
or click to select file
User Guide
Pick one PDF
Add a single file of 10 MB or less. Processing happens on a server, so the document is uploaded for the duration of the job and cleared afterwards. Keep genuinely confidential material off any server tool, including this one.
Convert
Press Convert to JSON. The document is parsed page by page and serialised, so the wait grows with length rather than with file size — a dense fifty-page statement takes longer than a graphics-heavy brochure of ten pages.
Inspect the shape first
Before writing any parsing code, open the download and read the top level. Knowing whether you are iterating pages that contain blocks, or blocks that carry page numbers, saves rewriting your loop twice. A quick jq keys or a browser JSON viewer answers it in seconds.
Write against the real data
Test your code on the actual output rather than an assumed structure, and include one awkward document in that test — a page with two columns, or a form with fields. Those are the cases that break a parser written against a clean example.
About the PDF to JSON Converter
This exists for one purpose: turning a document into something a program can loop over. Extracting a PDF as one flat string is easy and usually useless — you end up writing regular expressions against a wall of words, guessing where one field ends and the next begins.
What survives that a plain text dump loses
The converter emits JSON organised by page and by block, and critically it keeps the coordinates. That single detail changes what you can ask. With positions you can read the value sitting to the right of a known label, take only what appears in the top third of a page, or discard a repeated footer because of where it sits rather than by matching its wording. None of that is possible once the layout has been flattened into a string.
Why reading order is sometimes wrong, and why that is not a bug
A PDF stores its drawing instructions in whatever order the producing application emitted them. That order is frequently not the order a human reads in — a two-column layout may alternate between columns, and a text box added late in editing may be drawn last regardless of where it appears. The converter reports what the file contains. Sorting blocks by their coordinates usually restores the sequence you expected, which is exactly why the coordinates are worth having.
Write your parser defensively
PDFs in the wild are untidy in ways that are hard to anticipate. A number can arrive split across two fragments because the producer nudged the letter spacing. A label may or may not carry its trailing colon. Check that a key exists before reading it, fail loudly rather than silently when the shape is not what you expected, and test against one genuinely awkward document rather than only a clean example. Every pipeline built on document conversion needs to stitch and validate rather than trust.
Practical limits
One file per run, 10 MB maximum, processed on a server. Scanned pages need OCR first, since there are no characters in an image for the converter to serialise.
Frequently Asked Questions
Is the JSON structure guaranteed to stay the same?
Treat it as stable but not contractual. Write defensively — check that a key exists before reading it, and fail loudly rather than silently if the shape is not what you expected. That is good practice against any document converter.
Can I get coordinates for every word?
Positional detail is preserved at the level the parser works with, which is generally blocks and lines rather than individual glyphs. It is enough for label-and-value extraction and for filtering by region, which covers most practical needs.
Why is the reading order wrong on some pages?
Because a PDF stores drawing instructions in whatever order the producing application emitted them, which need not match how a human reads the page. Multi-column layouts and text boxes are the usual culprits. Sorting blocks by their coordinates often restores the order you expected.
Does it handle filled-in form fields?
Form values that were flattened into the page appear as ordinary text. Live, unflattened form fields are a separate structure inside the PDF and may not surface in the same way, so check a sample before relying on it.
What is the file size limit?
10 MB per document, one at a time. Larger files should be split first — the Split PDF tool runs in your browser and can break a long document into parts that fit.
Will a scanned PDF work?
Not without OCR. A scan holds images, so the converter finds no text to serialise and returns a structure describing pages with nothing in them.