Case study · Both
Managing AI-assisted publishing across multiple websites
How I built AI Publish around several websites, with separate brand voices, article and image preparation, review decisions and delivery checks.
In brief
Running several sites creates repeated decisions about voice, dates, images and delivery. I built AI Publish to connect those stages in one hub. AI prepares material, people can review it, and the workflow keeps uncertain publication or social results visible for follow-up.
Who this can help
An individual managing several brands or a small marketing team responsible for different business websites.
The problem
An article can be finished in the editor while its image, approval or social delivery still needs attention. With several business voices in play, the question becomes which step is complete for which site.
AI Publish grew from the publishing pipelines and spreadsheet coordination I used across my site network. I brought those steps into a shared article, site and calendar model, then developed the access and customer-facing features needed for broader commercial use.
The fit matters here. Each site has its own readers, voice and purpose, so I built the shared workflow around those differences. The aim was to carry the decisions through the process without flattening the brands into the same content.
What I built and how it works
Work starts with a site’s profile and voice. Topics and dates belong to that site, and generation steps prepare content for review. The calendar supports a weekly quantity, selected weekdays or an interval from an anchor date. A re-dating preview lets the operator inspect a change before applying it.
The hub publishes to WordPress and Astro through different connections. WordPress receives content through its publishing integration; Astro receives a repository change. The Astro workflow also checks for remote edits and target content requirements, because an existing website may have changed outside the hub.
Pictures have their own decision point. Generating a candidate doesn’t immediately replace the selected image. A person chooses the one to use, and the previous image is retained for reversal.
Social copy then goes through the selected provider. Delivery checks look at the targeted accounts individually. Partial or unreadable evidence stays unresolved so a person can investigate it. An accepted request and a delivered post are different events in that workflow.
Where AI and vibe coding helped
I used AI-assisted development and vibe coding to build the hub around the publishing work. During use, models also generate content and images. The calendar, site permissions and delivery checks are application logic around that AI, with their own responsibilities.
What people still control
Access is assigned per site, with levels for viewing and approval, editing, or publishing. Approval gates are optional and must be enabled for each target. On a gated site, the approval must match the content being sent, including a changed body on a live update.
The operator still checks facts, voice, pictures and the result on the destination. AI generates material within the process; editorial responsibility stays with people.
What the evidence supports
The working hub connects planning, content preparation, publication and delivery evidence. For someone managing several brands, that makes the outstanding work visible in the same operation as the article itself.
For a small team, the more useful detail may be site-specific responsibility: someone can review one business’s content without receiving universal publishing authority across the whole network.
Limits to keep in view
The system is in use across the site network and in commercial development. That doesn’t establish paid customer adoption or measured marketing returns. The consolidated Review screen is still being extended, and some publication handoffs need further live validation. Separate approval isn’t enabled universally.
Frequently asked questions
Can every website have its own voice?
AI Publish uses site profiles for voice and editorial rules. Each profile needs deliberate configuration and review; a shared hub does not mean shared wording.
Does every article require a separate approver?
Approval gates are optional per target. On a gated site, approval must remain current for the body being published or updated.
Does a successful social send prove every account received it?
No. The verifier checks targeted accounts individually. Incomplete evidence remains unresolved for a person to investigate.
A useful next step
Bring me one article’s journey from idea to website and social post. I can help identify the decisions and checks that belong in your process. Talk with me about your publishing workflow.