What I measure, written out in full.
A site selling citability has to produce citable material. Here I publish the operational definitions and the analysis I use in the work: how measurement is done, what it can prove, and what it cannot. No promotional content dressed up as a deep dive.
One definitional piece a month, plus public analysis whenever there is data worth showing.
RSS feed- [ Definition ]
What makes a product page readable by a shopping agent
A shopping agent does not weigh persuasion, it verifies constraints. Price, availability, delivery time, returns, and variant attributes must be resolvable without interpretation. If a constraint cannot be verified, the page is dropped in favour of one that exposes it, even when the product is worse.
Reading: 7 min
- [ Definition ]
How to measure visibility inside a generative engine
You do not measure a ranking, you measure a frequency. Fix a question set, run each question several times in clean sessions, and record how often the brand appears and which sources the answer used. The useful number is frequency under stated conditions, compared against the same measurement repeated later.
Reading: 8 min
- [ Definition ]
GEO and AEO: what actually changes in practice
GEO and AEO are not synonyms. AEO works on extractability: making an answer liftable from a page. GEO works on citability: getting a generative engine to pick you as a source while it builds the answer. The first is a content-form problem, the second one of authority and entity consistency.
Reading: 8 min
- [ Definition ]
llms.txt: what it does, what it does not, and why I publish one anyway
llms.txt is a markdown file indexing a site's canonical pages. No major vendor uses it as a visibility signal, and a study across three hundred thousand domains finds no citation lift attributable to the file. It stays useful as an entity-consistency artifact, at almost no cost. That is why I publish one.
Reading: 8 min
- [ Definition ]
Entity clarity: getting recognised before getting cited
Before deciding whether to cite you, an engine has to work out who you are and tell you apart from namesakes. It cross-checks your structured data, your canonical identifiers, and how third-party sources describe you. When those three layers disagree the system does not pick the likeliest one: it stays uncertain and moves to a clearer source.
Reading: 8 min
- [ Definition ]
Why many AI crawlers cannot see your site
OpenAI's, Anthropic's and Perplexity's crawlers do not execute JavaScript: they read the HTML the server returns on the first request and move on. Googlebot, Gemini and Applebot do render. If price, copy or navigation only appear after rendering, the page is empty to the first group, and checking takes ten minutes.
Reading: 8 min
- [ Definition ]
Structured data for AI: what helps and what is theatre
Structured data declares what a page is about, it does not make the page rank. Google documents a finite list of types that produce a rich result, and states that correct markup still guarantees nothing. Anything outside that list is valid schema that triggers no feature.
Reading: 8 min
- [ Definition ]
How a generative engine builds an answer
A generative answer moves through four stages: interpreting the question, retrieving documents, checking that the retrieved text supports the claim, and synthesising. Citation is a separate fifth step, decided afterwards. Being retrieved is not being cited, and the two problems are fixed differently.
Reading: 9 min
How visibility inside a generative engine gets measured, written out in full. Read the article
Back home