WordGPT

Turn a PDF into structured JSON, one record per row

Paste text or upload a PDF to get its records as JSON, with names, figures and dates kept as written. Free preview; a free account gives the full file.

or drop a PDF, Word or text file here (up to 5 pages, 2 MB)

This preview shows the first 25 rows and does not keep your file.

Need a different result? Turn anything into anything: drop any file, pick what you want out of it.Try it →

A PDF stores text as positioned lines, not as data, so copying a table out of it often leaves names, amounts and dates jumbled together. This page reads the document and returns its records as JSON with up to eight fields each, keeping the values and section order as written. If the PDF does not give a field, it is left empty.

What you get

Records, not paragraphs

Tables and label-and-value lines become objects, with keys taken from the headings and labels in the PDF.

Values kept as written

Names, figures, dates and currency signs are copied exactly as they appear. Nothing is rounded or reformatted.

Empty means empty

If a field is not in the PDF, its key is left blank rather than guessed from nearby text.

How it works

  1. Step 1

    Paste the text from your PDF, or drop the file into the box.

  2. Step 2

    Check the table preview, which shows the records as rows and columns.

  3. Step 3

    A free account gives you the full JSON download, long and scanned files, batches of files, and the option to change the structure by asking.

Who uses it

  • Developers loading document data into an app

    Get clean objects to import into a database or API, with the values from the source left unchanged.

  • Bookkeepers and admins with supplier or order sheets

    Turn a printed order summary into structured lines that an import or accounting tool can read.

  • Researchers working with tables from reports

    Pull a results table from a report PDF into JSON so the figures can be processed in code.

Questions

What does the JSON look like?

Each record becomes an object, and each field uses the heading or label from the PDF as its key, up to eight fields per record. A table with the columns Item, Qty, Unit Price and Total becomes a list of objects with those four keys.

Will it change numbers, dates or names?

No. Figures, dates and names are copied as they appear in the PDF. If a value is missing, its field is left empty instead of being filled in.

Does the preview read the whole PDF?

No. The preview reads up to 5 pages of a PDF (2 MB) or about 6,000 characters of text, and it does not keep your file. A free account handles whole documents, long files and scanned PDFs, which need OCR.

Can the JSON contain mistakes?

Yes. Text copied out of a PDF can break across lines, so a record may split or a value may be misread. Check the JSON against the original when the figures or dates matter.

Related tools

See it on real examples

Three actual runs of this tool. Open one to see the input and the whole result.