Ghost and AI: Fully Automating 200 Articles with Step-by-Step Workflow and Tool List
AI Summary · From a Serial Entrepreneur’s Perspective (The following content is distilled by AI; viewpoints belong to the original author. Reading this summary is enough—skip the full article if you prefer.)
Xiangyu built a blogging workflow on Ghost and automated everything from topic selection to publishing using 45 CLI commands and 14 sub-workflows. Humans only handle the final review. Data points: 200+ articles published, with operational time compressed to less than one-fifth of what it used to be. This setup is ideal for developers, instructors, or companies that produce content regularly. It targets the pain of repetitive work and lifts human efficiency.
- Get Ghost automated workflows: 45 CLI commands cover create, read, update, and delete…
- Copied a three-layer review mechanism: AI quality check, mock readers nitpicking, human final say…
- Migration pitfalls to avoid: WordPress plugin conflicts, difficult AI read/write APIs…
- Set up a publishing pipeline: pick topics, research first-party sources, auto-insert images, schedule sends…
- Launch at zero cost: Ghost is open-source and free, with built-in membership subscriptions and email broadcasting…
Let machines handle the repetitive labor; humans focus only on the creative part
Independent developers, freelancers, and entrepreneurs who run blogs or websites face the same trap every day: find a topic, dig up research, draft, revise three times, add images, format, publish, and check links. Going from an idea to a live post takes at least half a day. Publishing three posts a week is already the limit.
Xiangyu (hh) flipped this script with an automated workflow built on Ghost. His site has published over 200 articles, and throughout the entire flow—from topic selection to launch—the human only steps in for the final approval. This isn’t a theoretical design; it’s a production line that already runs.
Why Ghost instead of WordPress?
When automating content, the real need isn’t whether the admin dashboard is pretty, but whether AI can read and write articles smoothly. WordPress has a massive user base, but it has three hard limitations in automation scenarios:
Heavy plugin dependency: Features are bolted on via plugins. After installing a dozen, conflicts are routine, causing white screens and broken functions.
Updates break things: If core, themes, and plugins aren’t version-aligned, the whole site can go down.
Poor AI compatibility: Articles are stored as a mix of HTML and shortcodes. For AI to cleanly read or write an article, it must wrestle with a tangle of legacy formats.
Ghost is built for automation from the start: open-source and free, it stores articles as structured data (a content tree), separates read/write permissions, and uses auth tokens that expire in five minutes. This lets AI target a specific paragraph or image for edits without breaking the whole structure, minimizing the risk of collateral damage.
Core architecture: 45 commands + 14 sub-workflows
The system has three layers. You can run them independently or chain them into a fully automated pipeline.
Layer 1: One-click site setup. A single deployment script runs on a blank server and spins up a blog with a domain, SSL security, and a Chinese theme in three minutes. If any config is missing, the script throws an error and halts, preventing rookie mistakes from reaching production.
Layer 2: 45 CLI management commands. This is the foundation of the entire system—one file, zero dependencies, covering every admin operation:
- Create, read, update, delete: Manage articles, tags, and members. Anything you’d click in the backend, you can now do with a single command.
- Scheduled publishing: Set exact publish times. Write a batch on weekends; let the system fire posts automatically on weekdays.
- Post-publish quality check: Automatically detect broken links, missing images, and formatting drift, plugging the “publish and forget” gap.
- Full-site sync check: Compare local drafts line by line against live articles, making any divergence or accidental edit immediately visible.
- Format import and optimization: Import plain text and auto-convert it to Ghost format (15 types supported), compress images, and upload them to cloud storage.
Layer 3: 14 sub-workflow pipelines. Feed the AI a topic, and it runs the complete loop from selection to publication.
Three-layer review mechanism: where AI ends and humans begin
The most valuable part of the pipeline is the “three-layer review” mechanism, which is also the key defense separating “AI-managed blogs” from “AI spam.”
- Machine quality check: Verifies compliance, readability, factual accuracy, and SEO elements. Fail the hard metrics, and the draft gets sent back for重写.
- Mock reader nitpicking: Introduces eight different virtual reader personas for multi-dimensional scrutiny, up to four rounds of back-and-forth. If the same issue stalls the process for two consecutive rounds, the system auto-stops to avoid spinning wheels.
- Human final approval: The first two stages are AI-led; the final stage returns the wheel to a human. The reviewer just scans the final draft, types “approve” when satisfied, and hits publish.
Throughout this process, one hard rule is worth borrowing for any content creator:Numbers must have a primary source. If you can’t verify it, remove it or downgrade to “roughly.” Never fabricate.
Actual results and launch advice
In Xiangyu’s own testing, after migrating to Ghost and running this workflow, content operations time shrank to less than one-fifth of the original. Over 200 articles now stay perfectly in sync between local drafts and the live site.
For readers who want to replicate this setup:
- Cold-start suggestion: Deploy the Ghost environment first—you only need a low-cost VPS. Focus on configuring the 45 CLI commands; they’re the foundation for connecting AI later.
- Risk warning: Ghost’s theme ecosystem is rich, but the technical bar is slightly higher than WordPress’s. If you lack hands-on skills, the initial build may take time. Also, AI-generated content must pass the “factual accuracy” gate strictly; otherwise, the first layer of the three-layer review will fail easily.
This approach proves one thing: the core competitiveness of content entrepreneurship is shifting from “the efficiency of repetitive labor” to “topic direction and final taste.” Hand over everything machines can do to machines, and humans can finally focus on the truly scarce creative work.
Original · Xiangyu Workflow:Read original →
Related tool recommendation (sponsored):Lingxia Software 1Panel Panel and Halo Mall