Build a Content Bot: RSS to AI Caption to Auto-Publish

A content bot is one of the most useful small automations you can build: point it at an RSS feed, and every time something new appears, it writes a caption and posts it to your social accounts. No dashboard, no copy-paste, no forgetting. Here's the whole pipeline, with the real endpoints, and the one detail that quietly breaks most people's first attempt.

The Pipeline in Four Stages

Every RSS-to-social bot, no matter the language, is the same four stages in a loop:

1. Read. Fetch the RSS feed and parse out new items.
2. Dedupe. Skip anything you've already posted.
3. Write. Send each new item to an LLM to draft a caption.
4. Publish. POST the caption to Publora, which fans it out to every connected network.

Stages one, two, and three are your code. Stage four is a single API call, because Publora already holds your account connections and handles per-platform formatting.

Stage 1 & 2: Read the Feed, Skip What You've Seen

Deduplication is the part everyone gets wrong on the first try. Without it, the first time the feed updates, your bot re-posts every item in it. Keep a record of published item IDs, use the RSS guid or the item link, and check against it before doing anything else.

import feedparser, json, os

SEEN = "seen.json"
seen = set(json.load(open(SEEN))) if os.path.exists(SEEN) else set()

feed = feedparser.parse("https://example.com/feed.xml")
new_items = [e for e in feed.entries if e.id not in seen]

Stage 3: Turn the Item Into a Caption

Don't post the raw headline. Feed titles are written for a blog index, not a social feed. Hand the title and summary to an LLM with a prompt that carries your voice, and ask for something native to the platform.

# any LLM works; keep the prompt short and opinionated
prompt = f"""Write a one-line social post about this article.
No hashtags, no emoji, sound like a person, not a press release.

Title: {item.title}
Summary: {item.summary}"""

caption = call_your_llm(prompt)

If you want the captions to genuinely not read like a bot, our guide to posts that don't sound like AI is the prompt-shaping checklist we use.

Stage 4: Publish to Every Platform in One Call

This is where a publishing API earns its place. Instead of integrating with LinkedIn's API, then Instagram's, then Threads', you connect each account to Publora once and make one call per post.

import requests

requests.post(
  "https://api.publora.com/api/v1/create-post",
  headers={"x-publora-key": os.environ["PUBLORA_KEY"]},
  json={
    "content": caption,
    "platforms": ["linkedin", "threads"],
    "scheduledTime": "2026-07-01T14:00:00Z"  # omit to post now
  }
)

seen.add(item.id)
json.dump(list(seen), open(SEEN, "w"))

Record the item as seen only after the post succeeds, so a failed request retries next run instead of silently vanishing. For posting images from the feed, request an upload URL first, push the file, then reference it, the same media flow covered in automating social media with Python.

Running It on a Schedule

The loop above runs once. To make it a bot, run it every 15 to 60 minutes. Three common ways:

Host	How it runs	Good for
cron on a server	A crontab line calls the script	You already have a box running
GitHub Actions	Scheduled workflow, free minutes	No server, code already on GitHub
Serverless (Lambda, Cloud Functions)	Timer trigger invokes the function	Zero maintenance, pay per run

Whichever you pick, keep the seen record somewhere persistent, a small database or object store, not local disk that resets between runs.

Knowing It Worked

A bot you can't see is a bot you don't trust. Two options: poll Publora's list endpoint to confirm a post shipped, or register a webhook so Publora tells your app when a post publishes or fails. Webhooks are the cleaner choice, no polling, and you find out about failures immediately. The mechanics are in rate limits, tokens, and webhooks.

FAQ

How do I auto-publish an RSS feed to social media?

Poll the feed on a schedule, skip items you've already posted, send each new item to an LLM for a caption, and POST that caption to Publora's create-post endpoint with your chosen platforms. A cron job or serverless function runs the loop.

Do I need a separate integration per platform?

No. Connect each account to Publora once, then one create-post call fans the content out to LinkedIn, Threads, Instagram, and the other supported networks.

How do I stop it posting duplicates?

Keep a record of item IDs (the RSS guid or link) you've already published, and skip any item already in it. Store that record somewhere persistent, not disk that resets between runs.

Further Reading

• Automate Social Media Posts with Python + Publora
• What Is a Social Media Scheduling API?
• Write Social Media Posts That Don't Sound Like AI

---

About the author. Written by the Publora team. Code is illustrative; the Publora endpoints shown reflect the live REST API documented at docs.publora.com.