How to Extract Structured Data from PDFs with n8n

Copying fields out of PDFs by hand does not scale. This guide builds an n8n workflow that pulls the text from a PDF, hands it to a language model with a fixed schema, and writes the result as structured JSON into Google Sheets or a database. The whole thing is four nodes plus a chat model, and it runs unattended once it is on.

• An n8n instance (cloud or self-hosted)
• A source for the PDF, such as a Google Drive folder, an email inbox, or an HTTP URL
• An OpenAI or Anthropic API key for the chat model
• A destination: a Google Sheet, Postgres, or any database node

Step 1: Get the PDF into the workflow as binary

The text extraction node needs the actual file data, not a link. Start with a trigger, then add a node that returns the PDF as binary: a Google Drive node with the Download File operation, an HTTP Request node pointed at a file URL, or the Gmail node reading an attachment. Run it once so the PDF arrives on a binary property (n8n names it data by default).

Step 2: Pull the raw text with Extract From File

Add the Extract From File node and set Operation to Extract From PDF. Point Input Binary Field at the property that holds your file (data, unless you renamed it). On run, the node converts the PDF into JSON and puts the document text on an output field named text, which you will feed to the model in the next step.

Step 3: Extract structured fields with an LLM

Add the Information Extractor node from the Advanced AI section. It needs a chat model attached as a sub-node, so connect an OpenAI Chat Model or Anthropic Chat Model underneath it and add your API credential. In the Text field, map the extracted text with an expression such as {{ $json.text }}. Then choose how you describe the output shape.

{
  "vendor": "Vellum Co",
  "invoice_number": "INV-2041",
  "date": "2026-06-12",
  "total": 1240.00,
  "currency": "USD",
  "line_items": [
    { "description": "Design retainer", "amount": 1240.00 }
  ]
}

Step 4: Write the JSON to Sheets or a database

The Information Extractor outputs your fields under an output object, ready to map directly. Add a Google Sheets node with the Append Row operation, or a Postgres / MySQL Insert node, and map each field (for example {{ $json.output.vendor }} and {{ $json.output.total }}) to a column. No separate JSON parse step is needed because the node already returns structured data.

Result: any PDF that lands in your source now becomes a clean, schema-checked row in seconds. Test with a few real documents and compare the values against the originals, then activate the workflow so it runs on every new file.