The n8n HTML node, and why it's my favourite output trick.

I've been running n8n in production for a year now — for editorial automation, internal ops, and the AI-assisted content pipelines I build for clients. About six months in, I hit the same problem on three different workflows: the output looked like garbage every time it left the platform.

JSON with literal \nsequences. Markdown that got mangled in Notion. Plain-text descriptions where every "smart quote" turned into ’. The fix turned out to be simpler than I expected, and it's been my default move ever since.

The problem: JSON is a transport, not a presentation layer.

Most n8n workflows hand you data as JSON, which is great — until you need a human to read it. The default options are all bad in their own way:

• Plain text strips formatting and makes long descriptions a wall of run-on lines.
• Markdown works in some destinations (Notion, GitHub) and breaks in most (YouTube, Gmail, Slack rich-text).
• JSON-as-textis what you get if you don't transform anything — escape sequences and all.

The fix: the HTML node, used as a templating layer.

The HTML node in n8n is usually framed as "scrape a page" — but it's also a Handlebars-flavoured template renderer. You can pipe any upstream JSON into it and get formatted output the next node can consume. Here's the pattern I use:

// Handlebars template inside the HTML node
{{#each items}}
  <h3>{{title}}</h3>
  <p>{{description}}</p>
  <a href="{{url}}">Read more →</a>
{{/each}}

Set the node's output mode to HTML and the next node receives a clean, rendered string. From there I usually pipe it into one of four destinations:

Without vs. with the HTML node.

Where it earns its keep.

Quick FAQ.

Try this on your next workflow. Drop an HTML node between your data-shaping nodes and your delivery node, give it 5 minutes of template-writing time, and watch the output quality jump.