This automation reads a blog's RSS feed every morning, picks the five newest posts, asks an AI model to write one plain sentence about each, and emails you the result. RSS (Really Simple Syndication) is a public file most blogs publish that lists their latest posts with titles, links and short excerpts, so the workflow can read it without logging in anywhere.

It is for solo founders, small teams and creators who follow a handful of industry blogs, competitors or newsletters with web versions, and who lose time opening tabs to check what is new. Instead of skimming five pages, you read one email with five sentences and five links, then open only the posts that matter.

Time saved is an estimate: if checking your reading list takes 10 to 15 minutes a day, this cuts it to about 2 minutes, which is roughly an hour a week. Building it takes 20 to 30 minutes the first time.

What you need

  • n8n, the workflow tool we build in. You can use n8n Cloud or run the free, self-hosted version on your own server. As of 18 November 2026, the n8n pricing page lists Cloud plans from €20 a month (Starter, 2,500 workflow runs, billed annually) and a Cloud trial with no credit card. This workflow runs once a day, so it uses about 30 runs a month.

  • An OpenAI API key for the AI summary. OpenAI charges per token (a token is roughly three-quarters of a word). Our test run used about 734 tokens for five posts; check the current rate for your chosen model on the OpenAI API pricing page, as models and prices change often.

  • A Gmail account to send the email. Gmail is free; you connect it to n8n with Google sign-in (OAuth).

  • An RSS feed URL. Many blogs publish one at /rss, /feed or /rss.xml. We use the public n8n blog feed in the examples.

The flow at a glance

The workflow has one trigger and five steps, run left to right:

  1. Every morning at 7am (Schedule Trigger) starts the run once a day.

  2. Read the RSS feed (RSS Read) fetches every post listed in the feed.

  3. Keep the 5 newest posts (Limit) drops the rest.

  4. Combine posts into one list (Code) turns five separate items into one block of text.

  5. Summarise the posts with AI (Basic LLM Chain with an OpenAI Chat Model) writes the digest.

  6. Email me the digest (Gmail) sends it to you.

Illustration of the n8n canvas for the RSS digest workflow

Illustration of the n8n canvas, drawn from the built workflow. The OpenAI Chat Model hangs below the AI step because it is a sub-node: it supplies the model, and the chain above it does the work.

Each step in n8n is a node. Data moves between nodes as items, which are small records (one per blog post here). Keeping that in mind explains most of the settings below.

Build it step by step

Create a new workflow in n8n, name it something you will recognise later (we used "Summarise new blog posts from an RSS feed with OpenAI and email me"), then add the nodes below in order. The tables show the exact settings from our built and tested workflow. Fields not listed stay at their defaults.

1. Every morning at 7am (Schedule Trigger)

A trigger is the node that starts a workflow. The Schedule Trigger starts it at a set time, so nothing has to happen in another app first. Add it with the plus button, search for "Schedule", and set:

Field

Value

Trigger Interval

Days

Days Between Triggers

1

Trigger at Hour

7am

Trigger at Minute

0

The time follows the workflow's time zone, which you set under the workflow's Settings. If yours shows UTC, change it to your own city or the email will arrive at the wrong hour.

2. Read the RSS feed (RSS Read)

This node downloads the feed and outputs one item per post, each with fields such as title, link, contentSnippet (a plain-text excerpt) and isoDate (the publish date). It needs no account or credential.

Field

Value

URL

https://blog.n8n.io/rss/ (replace with any feed you follow)

Options

none

Click Execute step once. You should see a list of posts in the output panel. If it is empty, the URL is not a feed (see Common problems below).

3. Keep the 5 newest posts (Limit)

Feeds often hold 10 to 50 posts. The Limit node passes on only the first few items and drops the rest. Most feeds list the newest post first, so keeping the first five keeps the five newest.

Field

Value

Max Items

5

Keep

First Items

4. Combine posts into one list (Code)

At this point there are five items. If you sent them straight to the AI node, n8n would call the AI five times and you would get five emails. The Code node runs a short piece of JavaScript once and merges the five items into one item holding a numbered text list.

Field

Value

Mode

Run Once for All Items

Language

JavaScript

Paste this code:

const posts = $input.all().map((item, i) => {
  const post = item.json;
  const snippet = (post.contentSnippet || '').slice(0, 400);
  return (i + 1) + '. ' + post.title + '\n' + post.link + '\n' + snippet;
});
return [{ json: { postCount: posts.length, postList: posts.join('\n\n') } }];

Line by line:

  • $input.all() collects every item that arrived from the Limit node, here five posts.

  • .map((item, i) => { ... }) runs the code inside the braces once per post; i is the position, starting at 0.

  • item.json holds the post's fields (title, link and so on).

  • The snippet line takes the excerpt, or an empty string if a feed has none, and cuts it to 400 characters so the AI prompt stays short and cheap.

  • The return inside the map builds three lines per post: number and title, link, excerpt. The \n is a line break.

  • The last line returns a single item with two fields: postCount (how many posts) and postList (all posts joined with a blank line between them).

5. Summarise the posts with AI (Basic LLM Chain + OpenAI Chat Model)

The Basic LLM Chain sends one prompt to a large language model (LLM, the kind of AI behind ChatGPT) and returns the reply in a field called text. It needs a model attached as a sub-node: click the Model connector under the node and add OpenAI Chat Model.

Field

Value

Source for Prompt (User Message)

Define below

Require Specific Output Format

off

Chat Model (sub-node)

OpenAI Chat Model, credential: your OpenAI account, model: gpt-5-mini

Set the prompt as an expression (switch the field to Expression) and paste:

Below are {{ $json.postCount }} blog posts. For each post write three lines: the number and the title, then the link, then one plain sentence that says what the post is about. Leave a blank line between posts. After the last post add one line starting "Common thread:" that names what the posts have in common. Use British spelling, plain words and no hype. Write plain text only, no markdown.

{{ $json.postList }}

The parts in double curly brackets are expressions: n8n replaces {{ $json.postCount }} with the number from the Code node, and {{ $json.postList }} with the list itself. The instruction to write three lines per post matters. Our first version only asked the model to keep the number and link, and it dropped the titles; spelling out the format fixed it.

We picked a small model to keep the cost low. Any chat model you have access to will work; larger ones write slightly better sentences for a higher price per run.

6. Email me the digest (Gmail)

Add a Gmail node, choose Message and Send, and connect your Google account when asked. n8n opens a Google sign-in window and stores the connection as a credential.

Field

Value

Credential to connect with

Gmail account

Resource / Operation

Message / Send

To

[email protected] (your own address)

Subject (expression)

Your morning reading digest for {{ $now.toFormat("d MMMM yyyy") }}

Email Type

Text

Message (expression)

{{ $json.text }}

Options › Append n8n Attribution

off

$now is the current date and time; .toFormat("d MMMM yyyy") prints it as, for example, 19 November 2026. {{ $json.text }} is the summary from the AI step. Turning off the attribution removes the "sent automatically with n8n" line from the bottom of the email.

Test it

Leave the workflow inactive and click Execute workflow at the bottom of the canvas. This runs every node once, straight away, without waiting for 7am. Each node shows a green tick when it succeeds, and you can click any node to see what it produced.

Success looks like this: RSS Read shows a list of posts, Limit shows five, Code shows one item with postCount: 5, the AI step shows one text field, and an email lands in your inbox. In our test the four steps before Gmail ran live against the n8n blog feed and OpenAI; the AI step used about 734 tokens and took 12 seconds. The Gmail step ran on sample data, because our Gmail connection had expired and needed reconnecting (problem 1 below). Here is the start of the real digest the AI wrote:

1. AI made integrations the vendor's problem
https://blog.n8n.io/ai-made-integrations-the-vendors-problem/
It argues that AI and the rise of MCPs have moved the burden of integrations onto vendors and proposes enterprise-readiness as the right axis to judge tooling.

2. New in n8n: Public API rate limits that keep every team on your instance running
https://blog.n8n.io/new-in-n8n-public-api-rate-limits-that-keep-every-team-on-your-instance-running/
Describes new per-API-key public rate limits in n8n, including 429 retry headers and idempotency keys to support safe retries.

[Posts 3 to 5 follow in the same format.]

Common thread: practical, technical guidance on building, scaling and operating integrations and automation.

When you are happy with it, switch the workflow to Active (top right). From then on it runs every morning on its own.

Common problems and fixes

  1. Gmail step fails with "authorization grant is invalid, expired, revoked". This happened in our own build: the Google sign-in stored in n8n had expired. Open Credentials in n8n, open the Gmail credential and reconnect it. If you run n8n yourself with your own Google Cloud app, the n8n Google OAuth guide notes that with the app's publishing status set to Testing and user type External, consent and tokens expire after seven days, so expect to reconnect weekly until you change that status.

  2. The AI drops the titles or adds markdown. Small models follow loose instructions loosely. Describe the exact layout you want, line by line, as our final prompt does, and say "plain text only".

  3. You get five emails instead of one. The Code node is set to Run Once for Each Item, or it is missing. Set it to Run Once for All Items so it outputs a single item.

  4. RSS Read returns nothing or an error. The URL is a normal web page, not a feed. Open the URL in a browser; a feed shows raw XML (text full of angle-bracket tags). Try adding /feed or /rss, or look for the RSS icon on the blog. The RSS Read node documentation lists its one option, Ignore SSL Issues, for feeds with broken certificates.

  5. The email arrives at the wrong time. The Schedule Trigger uses the workflow's time zone, and falls back to the n8n instance's time zone if none is set, according to the Schedule Trigger documentation. Set it under the workflow's Settings.

Make it yours

  • Follow several blogs. Add a second RSS Read node from the same trigger, join both branches with a Merge node set to Append, then raise the Limit to 10. The Code node and prompt work unchanged.

  • Only include posts from the last day. Replace the Limit node with a Filter node that keeps items where isoDate is after {{ $now.minus(1, 'day') }}. On quiet days nothing passes the filter, so the workflow stops and no email is sent.

  • Send it somewhere else. Swap the Gmail node for Slack, Telegram or a Notion page if that is where you read. The message field stays {{ $json.text }}.

Download the workflow

Copy the JSON below, open a new workflow in n8n, and paste it onto the canvas (Ctrl+V or Cmd+V). n8n builds all seven nodes and their connections. Then open the OpenAI Chat Model and Gmail nodes and choose your own credentials, put your address in the To field, and change the feed URL. No credentials or account details are included.

{
  "name": "Summarise new blog posts from an RSS feed with OpenAI and email me",
  "nodes": [
    {"name": "Every morning at 7am", "type": "n8n-nodes-base.scheduleTrigger", "typeVersion": 1.3, "position": [240, 300], "parameters": {"rule": {"interval": [{"field": "days", "daysInterval": 1, "triggerAtHour": 7, "triggerAtMinute": 0}]}}},
    {"name": "Read the RSS feed", "type": "n8n-nodes-base.rssFeedRead", "typeVersion": 1.2, "position": [480, 300], "parameters": {"url": "https://blog.n8n.io/rss/", "options": {}}},
    {"name": "Keep the 5 newest posts", "type": "n8n-nodes-base.limit", "typeVersion": 1, "position": [720, 300], "parameters": {"maxItems": 5, "keep": "firstItems"}},
    {"name": "Combine posts into one list", "type": "n8n-nodes-base.code", "typeVersion": 2, "position": [960, 300], "parameters": {"mode": "runOnceForAllItems", "language": "javaScript", "jsCode": "const posts = $input.all().map((item, i) => {\n  const post = item.json;\n  const snippet = (post.contentSnippet || '').slice(0, 400);\n  return (i + 1) + '. ' + post.title + '\\n' + post.link + '\\n' + snippet;\n});\nreturn [{ json: { postCount: posts.length, postList: posts.join('\\n\\n') } }];"}},
    {"name": "Summarise the posts with AI", "type": "@n8n/n8n-nodes-langchain.chainLlm", "typeVersion": 1.9, "position": [1200, 300], "parameters": {"promptType": "define", "text": "=Below are {{ $json.postCount }} blog posts. For each post write three lines: the number and the title, then the link, then one plain sentence that says what the post is about. Leave a blank line between posts. After the last post add one line starting \"Common thread:\" that names what the posts have in common. Use British spelling, plain words and no hype. Write plain text only, no markdown.\n\n{{ $json.postList }}"}},
    {"name": "OpenAI Chat Model", "type": "@n8n/n8n-nodes-langchain.lmChatOpenAi", "typeVersion": 1.3, "position": [1200, 520], "parameters": {"model": {"__rl": true, "mode": "list", "value": "gpt-5-mini"}, "options": {}}},
    {"name": "Email me the digest", "type": "n8n-nodes-base.gmail", "typeVersion": 2.2, "position": [1440, 300], "parameters": {"resource": "message", "operation": "send", "sendTo": "[email protected]", "subject": "=Your morning reading digest for {{ $now.toFormat(\"d MMMM yyyy\") }}", "emailType": "text", "message": "={{ $json.text }}", "options": {"appendAttribution": false}}}
  ],
  "connections": {"Every morning at 7am": {"main": [[{"node": "Read the RSS feed", "type": "main", "index": 0}]]}, "Read the RSS feed": {"main": [[{"node": "Keep the 5 newest posts", "type": "main", "index": 0}]]}, "Keep the 5 newest posts": {"main": [[{"node": "Combine posts into one list", "type": "main", "index": 0}]]}, "Combine posts into one list": {"main": [[{"node": "Summarise the posts with AI", "type": "main", "index": 0}]]}, "Summarise the posts with AI": {"main": [[{"node": "Email me the digest", "type": "main", "index": 0}]]}, "OpenAI Chat Model": {"ai_languageModel": [[{"node": "Summarise the posts with AI", "type": "ai_languageModel", "index": 0}]]}},
  "settings": {"executionOrder": "v1"}
}

Sources