How emreelbeyoglu Automated a 350-Post Ghost Blog with OpenClaw
Built by Anonymous (Ghost Forum) · · Source: Ghost Forum
Result: According to the author, the site grew from about 314 posts to 350+ and gained about 150 unique visitors.
What they built
Ghost forum user emreelbeyoglu runs meaninginadream.com, a dream interpretation blog on Ghost that they describe as an SEO experiment. In a post from March 2026, they explain how they connected OpenClaw to the Ghost Admin API, search data, image generation and social media so the blog updates and grows on a schedule. They describe it as fully automated and hands-off, and they mention no manual review step.
How it works
According to the post, OpenClaw runs four cron jobs:
- Content Improver (every hour): picks the post that was updated longest ago, pulls People Also Ask questions from Google through the Serper API, generates 2 images, and rewrites the post with a short answer at the top, an FAQ built from the search data, internal links and comparison tables. It pushes the update through the Ghost Admin API and submits the URL to IndexNow and the Google Indexing API.
- New Post Creator (every 6 hours): finds a very specific dream topic the site doesn't cover yet, checks Google competition through Serper, generates 4 images, writes a 1,500 to 2,000 word post, publishes it to Ghost and submits it for indexing.
- Pinterest Autopinner (every 3 hours): spots new or unpinned posts, writes a pin description with hashtags, and creates the pin on a board chosen by the post's tags, using the Postiz CLI.
- X Autoposter (every 3 hours): uses the same detection logic, writes a tweet-length hook and posts it with the feature image.
The stack they list is Ghost for the CMS and hosting, OpenClaw for scheduling, tool access and Telegram notifications, Grok 4 for writing and editing, Ideogram v3 for images, Serper for search data, and Postiz for Pinterest and X. The post doesn't say where OpenClaw is hosted.
They also share a few lessons. Converting Ideogram's JPGs to WebP took images from about 1.2MB down to roughly 60 to 120KB. They prefer real People Also Ask data over invented FAQs. And they say long agent prompts with inline code made the model leak raw API output, so they keep prompts short and mention moving logic into helpers.
The result
The author says the site went from about 314 posts to more than 350 and gained about 150 unique visitors. They also say updated posts get re-indexed within hours.
The replies were not friendly. One commenter called it an "AI slop pipeline", and another pointed out that Google's Indexing API is officially limited to job postings and livestreams, so it may not speed up indexing for regular articles. That commenter also suggested manually reviewing the AI-written interpretations for accuracy.
How to build something similar
- Start with one scheduled job, such as refreshing your oldest post, and get it reliable before adding publishing or social jobs.
- Feed the agent real search data for questions and competition checks instead of letting it make up FAQs.
- Keep prompts short and put API calls and image conversion in small helper scripts the agent can run.
- Add a review step before posts go live, at least at first, so you can catch errors and thin content. Search engines and readers both punish low-quality pages.
The original post
Want to build something like this?
Join the Shipping Skool community free. Paid members get live bootcamp calls every week.
Join Shipping Skool — FreeMore use cases
- How Raphael Aubry Makes Motion Videos with Opus 5.5 in Claude CodeContent Creation
- How SageCut Uses a Decision Model to Cut Opus 5.5 Video Review CostsContent Creation
- How Scotty Beam Runs X Content with Seven Grok Bot Agents and PostizContent Creation
- How jzOcb Built a Writing Style Skill That Learns from Your EditsContent Creation