2026-08-24
No CMS does not mean no content model
Teams that drop the CMS usually drop content modeling with it. The schema was the discipline, not the tool - here is what a strict model looks like in files.
Written in public.
What I've learned, sourced and dated
Most posts start from a study, a dataset, or a test I ran, sourced in the sentence that makes the claim and corrected in public when I get it wrong. Start with the content model below - it is the closest thing here to how I actually work.
Content architecture
Deciding how content is shaped
For deciding what a page is made of before anyone writes a line.
Delivery and maintenance
Deciding how pages get built and served
For deciding where rendering happens and what that costs you later.
Working with a consultant
Deciding who to hire
For deciding whether the person quoting you is worth the money.
Evaluating AI-search claims
Deciding what to believe about AI search
For checking a claim someone made about getting your site cited by AI engines.
2026-08-16
llms.txt will not get you cited. Here is the data.
97% of published llms.txt files got zero requests in one measured month across 137,210 domains. Here's what the file is actually for.
2026-08-16
Structured data for AI search - what still matters in 2026
No AI system tested surfaced a fact that existed only in JSON-LD. FAQPage is dead. Here's the controlled evidence on what schema.org markup still does.
2026-08-16
ChatGPT, Perplexity and Claude cite almost entirely different sources
Two AI engines share roughly 9% to 25% of the domains they cite for the same query set. There is no single page that wins everywhere.
Something wrong with your own site?
Reading is free. If your site is slow or breaks on every change, the website audit finds the cause and prices every fix.