A Practical Workflow for AI-Written Blog Posts That Still Stay Factual

AI can draft useful blog posts, but the publishing system needs frontmatter, draft controls, validation, source notes, and Git-based review.

Why use MDX instead of a CMS?

MDX is a strong fit when the writer wants to draft in AI tools, review locally, and publish through GitHub.

It keeps posts portable as text files. It also allows reusable components like AnswerBox, KeyTakeaways, Definition, SourceNote, and FAQ without adopting a heavier publishing system.

What should the frontmatter include?

Every post needs enough metadata for search engines, RSS readers, social previews, and AI indexing.

The required fields on this site are title, description, excerpt, published date, updated date, category, tags, featured status, draft status, and canonical URL.

Frontmatter

Frontmatter is the structured metadata at the top of a Markdown or MDX file. It tells the site how to classify, display, index, and publish the post.

Why should AI drafts start unpublished?

AI models can write quickly, but they can also overstate, blend sources, or create clean-sounding claims that are not true.

The default should therefore be draft: true. A human review changes that value only after checking facts, links, claims, dates, and tone.

What does the publish flow look like?

Ask an AI model to draft the post using the prompt in BLOG_AUTHORING.md.
Save the result as content/blog/your-slug.mdx.
Keep draft: true while reviewing.
Run content validation and the local build.
Change to draft: false.
Commit and push to GitHub.
Let Vercel build and publish the site.

This workflow is implemented directly in the jalalshams.tech codebase through MDX content files, build-time validation, RSS, sitemap generation, and AI-readable llms.txt endpoints.