Workflow
Using AI in the Website Build Workflow
Most of the value is in the workshop, before a single public request ever touches a model API.
Discovery and rewrite
Models speed up the boring middle: clustering notes, proposing sitemap labels, and turning a transcript into a punch list. A person still runs the interview and still owns the scope.
We keep source notes in the project so the draft can be traced.
Design and engineering assists
A model can propose component states, empty-state copy, and test ideas. It can also suggest a pattern that fights the design system.
We accept the diff that matches the system and reject the one that adds a framework to save twenty minutes.
Decisions stay written down
Which model was used, what it was allowed to see, and what a human changed should be recoverable. Chat logs inside a personal account are not a project record.
The repository, not the model vendor, is the system of record.
Continue
Answer Engines and Website Structure
How to structure a site so people and answer engines can use it, without writing for bots or inventing rankings.
Read →Privacy and AI Website Integrations
What should never be pasted into GPT, Gemini, Grok, or Claude, and how live AI features stay limited to necessary data.
Read →When a Website Should Call an AI API
A decision test for live calls to GPT, Gemini, Grok, or Claude versus keeping the model in the build and shipping static HTML.
Read →