Resources · n8n integration

n8n Information Extractor nodeConfigure Information Extractor in n8n.

Free text goes in, named fields come out. The n8n Information Extractor node reads a message, a CV or a scraped page and returns the same JSON keys on every run. It exposes 3 parameters, 3 ways to describe the output shape, and needs one chat model plugged into its port.

Try the Information Extractor node on n8nSee the 3 parameters

Verified Trustpilot reviews · AI, automation & growth agency

Why automate

What does the n8n Information Extractor node do?

It takes a block of text and hands back a structured object whose keys you decide in advance. You describe the shape once, either as a list of attributes, an example JSON object or a full JSON schema, and every item that passes through comes out with those same keys. The reasoning is done by a chat model, a sub-node you connect to the node rather than configure inside it.

Three jobs cover most of what people build with it. The first is inbox triage: an email body lands in Text, the node pulls out company name, budget and deadline, and the result goes straight into a Google Sheets row without anyone reading the thread. Field names stay stable, so the spreadsheet columns never drift.

The second is document parsing. A CV or an invoice arrives as text from an earlier node, Schema Type is set to fromJson, and you paste one example object with the keys you care about. n8n reads the property names and types from that example and ignores the values. Handy when you already have a sample payload and do not want to hand write a schema.

The third is cleaning up what a model said in prose. A summarizer or an AI Agent returns a paragraph, and this node turns that paragraph into fields a downstream branch can test with an If. That is the difference worth knowing: an agent decides what to do next, this node only decides what the data looks like.

Pick a different node when the text already has a machine readable structure. A JSON payload from an API does not need a model to be parsed, and a fixed pattern like an order reference is cheaper to catch with an expression. Reach for the extractor when the wording changes from one item to the next and a regex would break.

Two limits are worth planning around. Extraction quality depends entirely on the chat model you connect, whether that is OpenAI, Google Gemini or another provider, and that provider bills its own API calls. And a schema generated from a JSON example marks every field as mandatory, which matters when some documents simply do not carry the value. If you are still comparing platforms before you build, the n8n review goes through the trade-offs.

Connections

What do you plug into the node?

n8n splits AI work between a root node, the one sitting in your main flow, and sub-nodes that hang off its ports underneath, one type of sub-node per port.

Ports (what plugs in)

  • Chat Modelai_languageModelrequired
01Chat ModelThis port carries the model that actually reads the text and fills the schema. Drag a sub-node onto it and the model brings its own provider credential, which is why the extractor itself asks for nothing.

This port carries the model that actually reads the text and fills the schema. Drag a sub-node onto it and the model brings its own provider credential, which is why the extractor itself asks for nothing.

Key parameters

  • Required: leave the port empty and the node refuses to run, since there is nothing to do the extraction.
  • OpenAI Chat Model or Anthropic Chat Model: the usual starting points when the provider account already exists.
  • Google Gemini Chat Model: another hosted option, connected the same way.
  • Ollama Chat Model: for a model running on your own machine, on a self-hosted instance.
Use cases
one extractor per workflow branch, each with its own model sub-node, so a high volume triage branch and a careful document branch can use different models without touching the schema.
Parameters

How do you set up each parameter?

The Information Extractor node has 3 parameters. For each one: the node as you configure it in n8n, what the parameter changes, and our field notes.

01

Text

text

What you see in n8n

Notes & use cases

Tells the node which text to read. Almost always an expression pointing at a field of the incoming item rather than something typed by hand.

Key parameters

  • Text: the text to extract information from, typically {{ $json.chatInput }} after a chat trigger or {{ $json.text }} after an Extract from PDF node. An expression is a small piece of code n8n evaluates per item.
Use cases
feed it the body of a support ticket so the next node gets a product name and a severity instead of a wall of prose.
02

Schema Type

schemaType

What you see in n8n

Notes & use cases

Decides how you describe the output shape. Each choice reveals a different field underneath.

Key parameters

  • Schema Type: fromAttributes lists attributes with types and descriptions, fromJson builds the schema from a sample, manual takes the schema as written.
  • Attributes: the per field list behind From Attribute Descriptions, filled one entry at a time with Add Attribute.
  • JSON Example: the sample object behind Generate From JSON Example. Property names and types are read, values ignored, and every field is treated as mandatory.
  • Input Schema: the JSON Schema behind Define using JSON Schema. The $refs syntax is not supported.
Use cases
start on attributes for a handful of fields, move to a written schema once nesting appears.
03

Options

options

What you see in n8n

Notes & use cases

Holds the settings that stay hidden until you need them. Nothing here is required, and the node runs fine with the collection empty.

Key parameters

  • System Prompt Template: replaces the system prompt used for the extraction. n8n still appends its format specification instructions, so the output shape stays intact.
  • Batch Processing: batch processing options for rate limiting, opened with Add Batch Processing Option.
Use cases
a custom system prompt to spell out house rules, like how a date should be written, when the default wording keeps producing an inconsistent field.
Need help

Need help automating Information Extractor with n8n?

A person reads every message.

FAQ

Questions people ask next

01Is the Information Extractor node free in n8n?
Yes. It ships inside the n8n AI package, so there is nothing to install and nothing extra to pay on the n8n side, on n8n Cloud as well as on a self-hosted instance under the Community Edition and its Sustainable Use licence. A workflow behaves the same in both places. The one cost sits elsewhere: the chat model you connect belongs to a provider such as OpenAI, Anthropic, Google or Mistral, and that provider bills its own API calls under its own terms. n8n adds nothing on top of that.
02What do you need to make it work?
No account, no key, no credential on the node itself. The Information Extractor has no Authentication selector and nothing to configure on the connection side, which is unusual enough to be worth saying out loud. What it does need is a chat model sub-node wired to its Chat Model port, and that sub-node carries the provider credential. So the setup work happens once, on the model, and the extractor inherits it. Leave the port empty and the node stops, because there is no model to read the text.
03What are the limits of the n8n Information Extractor node?
It extracts, and that is all it does. No writing anywhere, no branching, no calling an API: those belong to the nodes after it. The page describes version 1 of the node, and an older workflow may show fewer options. Two behaviours catch people out. A schema generated from a JSON example treats every field as mandatory, which is awkward when a document genuinely lacks a value. And the Input Schema field does not support the $refs syntax, so a schema built from references has to be flattened before you paste it.
04What do you connect at minimum, and which model to start with?
One sub-node is enough: a chat model on the Chat Model port. A sub-node is a small node that never runs on its own, it only works through the root node it is attached to. If the account already exists, OpenAI Chat Model or Anthropic Chat Model are the shortest paths, and Google Gemini Chat Model connects the same way. On a self-hosted setup with no provider account, Ollama Chat Model runs the model locally. Start with whichever provider your team already pays for, then change it later without touching the schema.
05n8n or Make for structured extraction?
It depends on where the data has to live. Make is hosted by its vendor, with no self-hosting option, and charges per operation, which is predictable when volume is steady. n8n runs on n8n Cloud or on your own Docker or npm install, so text that must not leave your servers can be processed in place, and the cost model follows the hosting rather than the number of items. On the AI side, the sub-node approach means you swap a model by dragging a different one onto the port. Both build visually, so the real question is hosting, data control and cost shape.
Hack'celeration Lab

Get our weekly integration tips.

No spam. Unsubscribe anytime.

Same platform

Other docs you might need on this stack.

Browse all integrations