Document parsing: how it works, methods, software and tools

Document parsing is the process of turning a document, such as a PDF, a scan, a Word file or an email, into structured data that software can use: named fields, tables and text in the right reading order. A parser reads the file, works out its layout, finds the values you need, such as the invoice number or every transaction on a statement, and outputs them as JSON, CSV or records in another system.

What is document parsing?#

Parsing means analyzing input and giving it structure. Data formats like JSON already carry their structure (see data parsing). Documents don't, for three reasons:

• Digital PDFs have no structure

  A PDF stores characters at positions on a page, with no notion of "this is a table row" or "this is the total."

• Scans have no text

  A scan or photo is just an image; there's no text until OCR reads it.

• Layouts vary

  Every vendor's invoice and every bank's statement puts the same information in a different place.

A document parser has to rebuild structure from what's on the page.

How document parsing works#

Take a scanned bank statement that arrives as an email attachment. It's an image, so the parser runs OCR first to turn the picture into characters, a step a digital PDF skips straight past. Next it works out the layout, finding where the header sits and following the transaction table as it continues onto a second page without repeating its column headers. Then it compares the layout and wording against known types and classifies the file as a bank statement rather than an invoice or a pay stub.

From there the parser pulls out the account number, the statement period and the opening and closing balance as named fields, and joins every transaction row across both pages into one table. It also checks the numbers against each other, so a running balance that doesn't foot, or a page that's too blurry to read, gets flagged for a person instead of exported as-is.

Here is the result on a bank statement: account details, balances and every transaction come back as named fields and one table.

Rule-based, AI and LLM parsers#

Production platforms combine them: document AI models for known types, an LLM for a new layout, and rules for validation. Parsing for business data (the invoice total, checked and loaded into a system) and parsing for AI search and RAG (the whole document as clean text) are different jobs, so decide which one you're doing first.

Document parsing software and tools compared#

The tools fall into four groups. Docsumo is our product, so it's first: a cloud platform, not a library you run on your own servers. Details for the others come from each vendor's own site or documentation, checked in September 2026.

Docsumo reads bank statements, pay stubs, tax forms, ACORD forms and invoices; see Docsumo Document AI. Cloud prices are first-tier list prices and fall at higher volume.

A minimal parser in Python

A minimal table read from a digital PDF with pdfplumber:

import pdfplumber

with pdfplumber.open("statement.pdf") as pdf:
    for page in pdf.pages:
        for table in page.extract_tables():
            for row in table:
                print(row)

This works on clean, digital PDFs with ruled tables. Scans, borderless tables and tables that break across pages need more work. For OCR on scans, see our Tesseract guide.

Document parsing use cases#

• Accounts payable

  Invoice headers and line items for PO matching. See AP automation.

• Lending

  Bank statements, pay stubs and tax returns for underwriting and income verification.

• Insurance

  ACORD forms, loss runs and certificates of insurance.

• Commercial real estate

  Rent rolls and operating statements.

• Legal and procurement

  Contract dates, parties, renewal terms and clauses.

• Healthcare

  Intake forms, referrals and claim forms.

• Knowledge bases and AI assistants

  Policies, manuals and reports parsed for search and RAG.

Build or buy?#

Build with libraries when you have a few stable, digital layouts and engineers to maintain the code. Buy a platform when you have many senders, scans, tables that span pages, or values that drive decisions. The hidden cost of building is upkeep: every new layout is more rules. If the parsing has to run inside your own product, Docsumo's document extraction API returns the fields as JSON and gives your users a review screen.

• 99%99%field-level accuracy across 250+ document types
• 95%+95%+of documents processed straight through, without manual review

How to choose a document parser#

Vendor sample files are picked to look good. Test on 50 to 100 of your own documents, and agree before you start what counts as a wrong answer.

We'd rather you shopped by results than by label. Document parsing, document AI and intelligent document processing all describe the same job in different words, and none of those labels tells you how a tool handles your worst scan. Test candidates on the documents you actually receive, not on the term a vendor puts on its homepage.

• Test on your own documentsInclude your worst scans, multi-column pages and longest tables.
• Score field-level accuracy and straight-through rateA parser can return text for every page and still get key fields wrong.
• Check table handlingMulti-page, borderless and nested tables.
• Send it a document type you didn't plan forA flag or low confidence score is fine; a confident wrong answer is the dangerous outcome.
• Check validation and reviewCan you add rules and review exceptions against the source page?
• Test the whole data pathCheck that the output schema loads into your system, not just that the JSON looks right. See integrations.
• Model the cost at your real volumePer-page prices that look small at 1,000 pages a month look different at 100,000.
• Check securitySOC 2 Type 2, HIPAA and GDPR where your data needs them.

Start from the job, not the label#

Document parsing rebuilds structure from files that never carried it. Libraries and rules work for a few fixed layouts. Document AI adds validation and review for the many layouts a normal inbox produces. A parser built for RAG keeps a document as text instead of a spreadsheet. For the broader picture, read what is data extraction.

Book a demo to see Docsumo parse your own documents, or start a free trial.

Frequently asked questions#