Receipt & Invoice OCR - PDF & Image to JSON/CSV

Extract structured data from receipts and invoices with OCR: PDF, JPG or PNG in; clean JSON rows and a ledger-ready ledger.csv out. Date, supplier, tax, totals - with a confidence score on every row, ready for your ledger, automation pipeline or LLM/RAG stack.

Works on any English documents; extra depth for Australian books (GST and ABN aware, ABN checksum-verified). Batch as many document URLs as you like. $0.01 per document - no subscription.

Quick start

1. Paste direct links to your receipt/invoice files (PDF, JPG or PNG) into Document URLs. Use direct-download links - a Google Drive or Dropbox share page link returns a web page, not the file (both services offer a direct-download URL form).
2. Run. Each document becomes one row in the dataset, and ledger.csv lands in the run's key-value store ready to import.

Example output row

{
  "documentDate": "2026-03-14",
  "supplier": "RICHARDSON ELECTRICAL SUPPLIES",
  "abn": "41068031407",
  "abnChecksumValid": true,
  "gst": 39.45,
  "total": 433.95,
  "currency": "AUD",
  "status": "ok",
  "extractionMethod": "pdf-text",
  "sourceUrl": "https://.../invoice-4471.pdf"
}

What it does

• PDFs: extracts the text layer directly (fast, exact)
• Scanned PDFs (no text layer): pages are rendered and OCR'd automatically, with a confidence score on every row - low-confidence extractions are flagged review, never passed off as clean
• Images (JPG/PNG): OCR via Tesseract, with a confidence score on every row
• Extracts: document date, supplier name, ABN (verified against the real ABN checksum, not just the digit count), GST amount, total
• Knows Australian invoice layouts: on a TOTAL EX GST / TOTAL GST / TOTAL INC GST block it reports the inclusive total, and it will not mistake a "Total (inc GST)" figure for the GST
• Flags honestly instead of guessing: needs-ocr (scanned PDFs), review (no total found, or a GST figure that looks implausible against the total), error (with the reason)
• Outputs: dataset rows + optional combined ledger.csv in the key-value store

What it does NOT do

• No scraping, no logins, no third-party data. You supply your own documents, they are processed inside YOUR Apify account, and the developer never sees or stores them. Outputs live in your own storage and expire on your account's normal retention schedule.
• No guessing: a scan too poor to OCR reliably is flagged (needs-ocr or review), never silently invented.
• Heuristic extraction: check the review-flagged rows. Your books deserve your eyes - extracted figures are not financial, tax or accounting advice; verify before use.

Pricing

Pay-per-event: a few cents per document actually processed. Documents that fail or need OCR are not charged as processed. No subscription.

Terms of use

Process only documents you own or are authorised to process. Not affiliated with or endorsed by Xero, Intuit QuickBooks, or any accounting vendor named for compatibility.