How to convert bank statements to JSON: schema, methods and validation

To convert a bank statement to JSON, extract the account details, statement period, opening and closing balances, and every transaction from the PDF or scan, then write them into a structured JSON object with a transactions array. For digital PDFs from one bank, a PDF parser can work; for statements from many banks, or scans and phone photos, a document AI tool or API is the practical choice. Either way, validate the result: the opening balance plus credits minus debits must equal the closing balance.

This guide covers a sample JSON schema, the conversion methods compared, a step-by-step process and the checks that make the output safe to use.

Why convert bank statements to JSON#

• Apps and APIs speak JSON. Lending platforms, underwriting models, accounting tools and data pipelines ingest it directly.
• It keeps the structure. Account, period, balances and transactions stay together in one object.
• It's easy to validate. Balances and running totals can be checked in code before anything downstream uses them.
• It scales. Once statements are JSON, analysis and categorization can run automatically.

Typical uses: loan underwriting and income verification, cash flow analysis, bookkeeping and reconciliation, and feeding transaction data to analytics or machine learning models.

A sample bank statement JSON schema#

Field names differ between tools, but a good output has the same parts. This one holds the statement shown above:

{
  "document_type": "bank_statement",
  "bank_name": "First Midwest Bank",
  "account_holder": "Harbor Street Bakery LLC",
  "account_number_last4": "4821",
  "account_type": "business_checking",
  "currency": "USD",
  "period_start": "2026-07-01",
  "period_end": "2026-07-31",
  "opening_balance": 42180.55,
  "closing_balance": 38906.21,
  "total_credits": 88412.10,
  "total_debits": 91686.44,
  "transaction_count": 142,
  "transactions": [
    {
      "date": "2026-07-02",
      "description": "ACH DEPOSIT SQUARE INC",
      "debit": null,
      "credit": 4812.30,
      "balance": 46992.85,
      "category": "card_processor_payout",
      "page": 1
    },
    {
      "date": "2026-07-05",
      "description": "PAYROLL GUSTO",
      "debit": 6420.00,
      "credit": null,
      "balance": 40572.85,
      "category": "payroll",
      "page": 1
    }
  ]
}

Design choices that save trouble later:

• Numbers, not strings, for amounts and balances, with separate debit and credit fields or one signed amount, never both.
• ISO 8601 dates (YYYY-MM-DD), because statements print dates in many formats.
• Masked account numbers unless you truly need the full number.
• Page reference for each transaction, so a reviewer can find it on the source.
• Confidence scores per field if your tool provides them.

Play video: Bank statement extraction using Docsumo AI [A Quick Guide]

Conversion methods compared#

For a handful of statements from one bank, a parser is fine. For a lending or fintech product that receives statements from hundreds of banks, a document AI API is the practical approach.

How to convert bank statements to JSON with document AI#

1. Send the fileUpload the PDF or image in the web app, email it to the tool's inbox, or post it to the API. Good tools split multi-account and multi-month files.
2. ExtractOCR and layout models find the header, summary and transaction table, joining tables that run across pages. Docsumo's bank statement extraction has 99% field-level accuracy across 250+ document types.
3. Review low-confidence valuesA reviewer sees each uncertain field highlighted on the page and corrects it.
4. ValidateRun the balance checks below automatically; Docsumo checks the running balances.
5. Receive JSONFetch it from the API, or receive it by webhook when processing finishes, then load it into your app or data store.

Docsumo's free trial covers 14 days and up to 1,000 pages, with API and webhook access; see pricing.

Validate the JSON before you use it#

These checks catch misread values, dropped rows and statements that don't add up:

• Balance reconciliation: opening_balance + total_credits − total_debits = closing_balance.
• Running balance chain: each transaction's balance = previous balance + credit − debit.
• Totals: the sum of transaction credits and debits equals the statement's summary totals.
• Dates: every transaction date falls within the statement period, in order.
• Continuity: each month's opening balance equals the prior month's closing balance.

A simple check in Python:

from decimal import Decimal

def reconciles(stmt):
    bal = Decimal(str(stmt["opening_balance"]))
    for t in stmt["transactions"]:
        bal += Decimal(str(t["credit"] or 0)) - Decimal(str(t["debit"] or 0))
        if t.get("balance") is not None and bal != Decimal(str(t["balance"])):
            return False
    return bal == Decimal(str(stmt["closing_balance"]))

Failures point to a misread value, a missing transaction or an altered statement. See how to spot fake bank statements.

Storing and securing bank statement JSON#

• Encrypt at rest and in transit, and restrict access by role.
• Keep the source file linked to the JSON for audit and disputes.
• Set retention rules so data isn't kept longer than your policy allows.
• Choose vendors carefully. Docsumo is SOC 2 Type 2, HIPAA and GDPR compliant.

To go further than conversion, read AI bank statement analysis or how to convert bank statements to Excel.

The bottom line#

Bank statement JSON is only as good as the extraction and checks behind it. Use a clear schema with numeric amounts and ISO dates, pick a method that handles the banks and scans you actually receive, and reconcile every statement before your app trusts it.

Book a demo with a few of your own statements, or start a free trial.

Frequently asked questions#