The shelf inside the model.

When a category query is put to ChatGPT, Gemini, Claude, or Perplexity, three or four houses get named. We engineer the conditions under which yours is one of them.

§ THE POSITION

AI search is not ranking. It is recall. A model recommends what it has been trained to regard as canonical. The work is to be treated as canonical — deliberately, repeatedly, and in public.

§ THE WORKINGS

Four instruments we run — continuously.

§ 01 · CITATION MEASUREMENT

We instrument the models.

A prompt library of 400-600 decision queries per engagement, run daily against every major LLM. We see exactly when you are cited, by whom, and in what position.

§ 02 · ENTITY CANONICALIZATION

We teach models who you are.

Wikidata, Wikipedia, Crunchbase, LinkedIn, reputable press. The canonical surfaces that LLM training pipelines treat as authoritative — aligned and maintained.

§ 03 · ANSWER-SURFACE TUNING

We rewrite what gets quoted.

When tracking shows a competitor is being cited where you should be, we engineer an answer that is provably more complete, balanced, and correctly formatted.

§ 04 · SHELF PLACEMENT

We defend the shortlist.

Models return three or four houses per category. We work — every week, without exception — to keep you one of them, on the queries that matter most to your revenue.

§ THE QUERIES · WHAT WE TRACK