Convert PDF to JSON
APIs and automations need objects, not a Markdown dump. With Smart Extraction you describe the JSON shape you want — flat fields plus repeating groups — extract from a digital PDF, correct anything uncertain, then download JSON.
How JSON extraction is used
Developers map invoice_number, po_number, and totals into downstream systems. Operators send line items as arrays instead of flattening tables by hand. Missing and Needs Review labels help you catch uncertain values before export. The flow is the same each time: define fields, extract, review, export JSON.
Before & After
Purchase Order PO-8891 Customer: Harbor Logistics Ship by: 2026-04-02 Item Qty Unit Line PALLET-WRAP 12 8.50 102.00 LABEL-ROLL 6 4.25 25.50 Order total 127.50
{
"po_number": "PO-8891",
"customer": "Harbor Logistics",
"ship_by": "2026-04-02",
"line_items": [
{ "item": "PALLET-WRAP", "qty": 12, "unit": 8.5, "line": 102 },
{ "item": "LABEL-ROLL", "qty": 6, "unit": 4.25, "line": 25.5 }
],
"order_total": 127.5
}Define the fields you need, review the results, then export JSON or CSV.
FAQ
Is the JSON schema fixed, or do I define it?
You define the fields and repeating groups. The export follows that schema after you review the result.
What happens when a value is missing?
Every requested field is labeled extracted, missing, or needs review. You can correct values in the review table before export.
Can I also download CSV from the same extraction?
Yes. After review you can export JSON, CSV, or Markdown. Export does not consume another page.
Does JSON export send the PDF to your servers?
The PDF file stays in the browser. Only extracted text is sent for AI processing. Export itself is generated in the browser from the reviewed result.
Are scanned PDFs supported?
The browser workflow requires a digital PDF with a text layer. The paid automation API also accepts scanned PDFs and photos and sends the file to the AI provider for reading.