A finished website is not a finished job
Betsy's site was done and it was good. The goal for this round was different: turn it into something that markets itself while she is out staging a house.
The recon changed the plan
The instinct with a content engine is to point it at the site and start producing. Dinko's constraint was blunter — no pages for the sake of pages — so a market recon ran before anything got written.
The data said to do less. The site did not need more pages; it needed the pages it already had pushed from the second page of results to the first. The content queue came out strengthen-heavy, and the single new page built that week was the one the data actually demanded: a vacation-rental design page for Bradenton, a market where the site was already ranking without having a page at all.
An AI drafts, a machine decides
Everything automated here follows the same rule: an AI writes the draft, a deterministic checklist decides whether it publishes, and the AI never touches anything live. Blog posts go out three mornings a week only if the checks pass. The Google profile gets a post every week from a library of sixteen. Her reviews sync themselves onto the site.
The point of the split is that a bad draft costs nothing. It fails the checklist and nobody sees it.
Introducing her to the machines
People increasingly ask an assistant for a recommendation before they ask a search engine. So the site carries a canonical brief written for those readers — who she is, where she works, what she does and does not do — plus a short version in the footer of every page. If a model is going to describe her business, it should describe it in her words.
