🧾 AI OCR for Tax Documents (Invoices, Balance Sheets & Tables)

Turn invoices, receipts, balance sheets, bank statements and tabular PDFs into clean, structured JSON with AI. Upload a file or pass document URLs, and get back the document type, issuer/recipient, dates, totals, taxes, line-item tables and a summary - ready for your accounting system or spreadsheet.

🎯 Built for Tax & accounting teams. Not a generic text dump: the AI detects the document type and extracts the fields that matter, plus the tables, preserving layout meaning.

---

What you get (per document)

Field	Description
documentType	invoice, receipt, balance_sheet, income_statement, bank_statement, purchase_order, table, other
issuerName / issuerTaxId	Vendor/company and tax ID (VAT, CNPJ, EIN...)
recipientName / recipientTaxId	Buyer/customer and tax ID
documentNumber, issueDate, dueDate	Document identification
currency, subtotal, taxAmount, totalAmount	Monetary fields (plain numbers)
tables[]	Extracted tables (line items, balances...) with columns + rows
keyValues	Any other labelled fields (payment terms, account no., period...)
summary	One-line description
fileMetadata	type, sizeBytes, pageCount (PDF)

---

How to use

Upload a file in the input, or pass URLs for batch:

{
  "documentUrls": [
    "https://example.com/invoice.pdf",
    "https://example.com/receipt.jpg"
  ]
}

Supports PDF, PNG, JPG and WebP. Up to 50 documents per run (send larger volumes via sequential calls). PDFs are read natively (multi-page); images are auto-optimized before analysis.

---

Pricing

Charged per document successfully extracted (event tax-document-extracted). Documents that fail to download or can't be read are not charged.

---

Example output

[
  {
    "documentUrl": "https://example.com/invoice.pdf",
    "success": true,
    "documentType": "invoice",
    "issuerName": "ACME Ltda",
    "issuerTaxId": "12345678000190",
    "recipientName": "Globex Inc",
    "documentNumber": "INV-2024-001",
    "issueDate": "2024-03-15",
    "dueDate": "2024-04-15",
    "currency": "USD",
    "subtotal": 1100.0,
    "taxAmount": 150.0,
    "totalAmount": 1250.0,
    "tables": [
      { "title": "Line items", "columns": ["description", "qty", "unitPrice", "total"],
        "rows": [ { "description": "Consulting", "qty": 10, "unitPrice": 110, "total": 1100 } ] }
    ],
    "keyValues": { "paymentTerms": "Net 30" },
    "summary": "Invoice from ACME Ltda to Globex Inc, total USD 1250.",
    "fileMetadata": { "type": "pdf", "sizeBytes": 84210, "pageCount": 1 },
    "failureReason": null,
    "processedAt": "2026-01-01T12:00:00.000Z",
    "error": null
  }
]

---

FAQ

Which documents work best? Clear digital PDFs and sharp scans/photos. Very low-resolution or handwritten documents may not be readable - the reason is reported in failureReason.

Does it handle multi-page PDFs? Yes. PDFs are read natively, including tables and layout, across pages.

Can I upload a file directly? Yes - use the upload field in the input, or call the API with a document URL.

Can I call it in real time? Yes. The Standby endpoint POST /extract responds synchronously. See below.

---

🔌 API integration

Batch run:

curl -X POST "https://api.apify.com/v2/acts/acme-ai~ocr-tax-document-ai/run-sync-get-dataset-items?token=YOUR_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"documentUrls":["https://example.com/invoice.pdf","https://example.com/receipt.jpg"]}'

Standby (POST /extract):

curl -X POST "https://acme-ai--ocr-tax-document-ai.apify.actor/extract" \
  -H "Authorization: Bearer YOUR_APIFY_TOKEN" \
  -H "Content-Type: application/json" \
  --compressed \
  -d '{"documentUrls":["https://example.com/invoice.pdf","https://example.com/receipt.jpg"]}'

The token goes in the Authorization: Bearer header, never in the URL.

---

Notes

This Actor analyzes documents you provide. You are responsible for having the right to process them and any personal or financial data they may contain.