How AI search actually picks brands, what a knowledge graph does that vectors can't, and what we find when we tear a brand down. Answer-first, one idea per paragraph, bylined to the engineer who builds the systems — written to be read by humans and cited by machines.
Each post opens with the answer, then earns it. If you only read the first paragraph, you should still leave knowing the thing — that is the format, and it is the same discipline we install for clients.
Not a vibes report and not an SEO scan. A week-long teardown that measures how AI search describes you versus competitors, names the architectural cause of the gap, and hands back a roadmap ranked by ROI. Here is what comes out the other end.
Plain RAG retrieves text that looks similar to the question. GraphRAG retrieves facts that are actually connected. For anyone whose product answers are relational — this dose, this claim, this contraindication — that difference is the whole ballgame.
Ask an assistant for the best brand in your category and you get a short, confident shortlist — the same few names, over and over. Here is the mechanism behind that, and why being absent from it is an architecture problem, not a marketing one.
More on the way. For the vocabulary used across these, see the glossary; for the free version of the diagnosis, the AI-Visibility Teardown.
Once the why is obvious, the $1,500 AI Brand Audit is the next step — your real share-of-model numbers, the architectural cause, and a roadmap you keep.