Build a AI Content Curator — LinkedIn Engagement + Weekly Newsletter from RSS Feeds Automatically Workflow with n8n

Every Monday morning, the same ritual eats an hour of your week: open a dozen RSS feeds, skim industry news, decide what's worth commenting on, write LinkedIn comments that don't sound like a bot, and then somehow turn all of it into a newsletter your audience actually opens. Do it consistently and you build authority. Skip two weeks and the algorithm forgets you exist. This article shows you how to hand the entire loop to n8n and GPT-4o — so the curation, the commentary, and the newsletter draft are waiting for you before your first coffee.

The problem: content curation is high-value work trapped in low-value labor

Thought leadership on LinkedIn and a steady newsletter are two of the highest-ROI channels for a technical founder or ops team. They compound. But the actual mechanics are brutally manual and easy to drop:

• Discovery is scattered. Your signal lives across 8–15 RSS feeds, and the good stuff is buried under press releases and rehashed news.
• Commentary doesn't scale. A generic "Great post!" comment does nothing. A sharp, on-topic comment in your voice earns profile visits — but writing five of them takes real cognitive effort.
• Newsletters die from inconsistency. The moment you miss a week, cadence breaks, open rates slide, and the list goes cold.

The trap is that this is strategic work — being visible in your industry — dressed up as clerical work: reading, copy-pasting, and reformatting. That mismatch is exactly what an automation layer plus an LLM is built to fix.

The solution: a weekly curation engine on n8n + GPT-4o

The workflow runs on a single trigger every Monday and produces three finished assets: a ranked shortlist of the week's five best articles, LinkedIn engagement comments written in your voice, and a newsletter draft ready to paste into Beehiiv or Mailchimp. Here's the pipeline end to end:

• Schedule Trigger fires Monday at 7:00 AM.
• RSS Read nodes pull the last week of items from every feed you track.
• A Code node dedupes and filters to the last seven days.
• GPT-4o scores and selects the top five articles against your criteria.
• A second GPT-4o pass writes a LinkedIn comment per article in your tone.
• A third pass assembles the newsletter (intro, five sections, CTA).
• Gmail / Slack delivers the whole package to you for a final glance before publishing.

The AI never publishes autonomously. It does the reading and the drafting; you keep the 60-second approval step. That's the right division of labor — leverage without losing your judgment or your voice.

Step-by-step setup in n8n

Assuming a self-hosted or cloud n8n instance and an OpenAI API key, here's how the nodes connect.

1. Schedule Trigger. Add a Schedule Trigger node, set the interval to Weeks, trigger on Monday, at hour 7. This is your only entry point — no manual runs needed.

2. Pull the feeds. Add one RSS Read node per feed URL (or drive a list through a Split In Batches loop if you track many). Each node returns items with title, link, contentSnippet, and isoDate. Wire them into a Merge node set to Append so everything lands in one collection.

3. Filter and dedupe. Add a Code node that keeps only items from the last 7 days and drops duplicate links:

const weekAgo = Date.now() - 7*24*60*60*1000;
const seen = new Set();
return items.filter(i => {
  const d = new Date(i.json.isoDate).getTime();
  if (d < weekAgo || seen.has(i.json.link)) return false;
  seen.add(i.json.link);
  return true;
});

4. Select the top 5. Add an OpenAI node (Chat model, gpt-4o). Pass the filtered titles and snippets as a JSON list in the user prompt. In the system prompt, define your selection criteria — for example: "You curate for a B2B RevOps audience. Rank by practical actionability and originality; reject press releases and funding news. Return exactly 5 items as JSON with title, link, and a one-line why." Set temperature to 0.3 for consistent ranking and enable JSON output so the response parses cleanly downstream.

5. Write the LinkedIn comments. Add a second OpenAI node. Feed it the five selected articles and a system prompt that carries your voice: tone, sentence length, whether you use emojis, your typical point of view. Instruct it to write a 2–3 sentence comment per article that adds a genuine insight or asks a sharp question — never a compliment with no substance. Bump temperature to 0.7 here so the comments read human. Output JSON keyed by article link.

6. Draft the newsletter. Add a third OpenAI node that receives the five articles plus their why lines and produces a full newsletter: a short hook intro, five sections (headline, 2–3 sentence summary, link), and a closing CTA. Ask for clean HTML output so it drops straight into Beehiiv or Mailchimp without reformatting.

7. Deliver for review. Add a Gmail (or Slack) node that sends you one message containing the newsletter HTML and the five LinkedIn comments. To push the draft directly into your ESP, use an HTTP Request node against the Beehiiv or Mailchimp API to create a draft campaign — draft, not send, so you always keep the final publish click.

Benefits: what you actually get back

• Time. A weekly ritual that took 60–90 minutes collapses into a 5-minute review. That's roughly 60 hours a year returned.
• Consistency. The newsletter ships every Monday whether or not you're inspired, traveling, or slammed. Cadence is what builds a list — the automation guarantees it.
• Compounding visibility. Five thoughtful LinkedIn comments a week, every week, is a legitimate top-of-funnel strategy. It puts you in front of other people's audiences without producing a single original post.
• Your voice, preserved. Because the tone lives in the system prompt, the output sounds like you — not like generic AICopy. Refine the prompt once and every future run inherits it.
• One source of truth. Add or remove a feed in one place and both your LinkedIn engagement and your newsletter update together.

Common pitfalls (and how to avoid them)

Letting the AI publish unsupervised. Don't wire the LinkedIn comment output straight into an auto-poster. LinkedIn actively penalizes automated engagement, and a tone-deaf comment on the wrong post costs you credibility. Keep the human approval step — that's the entire point of drafting versus publishing.

Vague selection criteria. If your system prompt just says "pick the best articles," GPT-4o will pick whatever is most popular, which is usually news you already saw. Be specific about audience, format, and what to reject. The quality of your curation is entirely a function of prompt precision.

Skipping deduplication. The same story often hits three feeds in slightly different wording. Without the dedupe step, your top five can collapse into two topics repeated. The Set-based filter above is cheap insurance.

Ignoring token and cost limits. Passing full article bodies from 15 feeds into GPT-4o gets expensive and can blow the context window. Send titles and snippets only for the selection pass; you don't need full text to rank relevance. Fetch full content only if you later add a summarization step.

No error handling on feeds. RSS endpoints go down. Set each RSS Read node to Continue On Fail so one dead feed doesn't kill the whole Monday run, and add an If node to alert you when fewer than five articles survive the filter.

Get these five right and the workflow runs untouched for months. You stop thinking about curation as a chore and start treating it as an owned channel that maintains itself — which is exactly what a busy technical team should demand from its tooling.

PRONTO PARA USAR

Ja construimos isso pra voce

Nao comece do zero. O AI Content Curator — LinkedIn Engagement + Weekly Newsletter from RSS Feeds Automatically e um workflow n8n pronto para instalar — conecta suas ferramentas em minutos, sem codigo.

Instalar por $199 →