What Gemini consults
Gemini, deployed across Google's consumer surfaces and inside the Workspace product line, is the model with the most direct access to Google's structured graphs. Its commercial-discovery answers are heavily anchored on entities the Knowledge Graph already knows about, on Google Business Profiles where local intent is detected, and on editorially-credentialed sources within Google's broader trust framework.
The implication is consequential. A subject that is poorly resolved inside Google's own systems is poorly resolved inside Gemini. The single highest-leverage move on Gemini is therefore the same as the single highest-leverage move on Google's broader presence: be canonical inside Google.
The Google graph integration
Three integration points are operationally material.
- Knowledge Graph entry. Where applicable, presence in the Knowledge Graph as a recognised entity transforms Gemini's behaviour toward the subject.
- Google Business Profile. For local-intent queries, Gemini draws heavily on the Profile, including reviews, attributes, and verified information.
- Editorial signals. Coverage in sources Google's broader trust framework recognises lifts both classic search and Gemini in parallel.
Gemini and AI Overview improvements correlate strongly. Subjects engaged on a brief that addresses Google's underlying graph see lift on both surfaces in approximately the same quarter. The work is, in operational terms, the same work.
The three moves that work
One — entity canonicalisation in Google's terms
Knowledge Panel where applicable, Google Business Profile reconciled to canonical entity records, schema appropriate to the entity type, sameAs assertions linking the subject to authoritative third-party records.
Two — editorial citation under Google's reranker
Coverage in sources Google's broader trust framework recognises — trade press, professional bodies, university and government domains. Gemini's behaviour follows Google's reranker; what lifts the reranker lifts Gemini.
Three — schema appropriate to the question
Question-type schema (FAQPage, HowTo, MedicalBusiness, etc.) accurately applied. Gemini, like the AI Overview, treats schema as a reranker confidence signal.
What does not work
- Approaches that treat Gemini as separate from Google's broader infrastructure.
- Pseudo-FAQ blocks designed for inclusion that do not anchor to the page's substance.
- Inclusion in low-editorial directories that Google's trust framework increasingly weights down.
- Schema markup that overstates the page's structure or relevance.
How the institution measures it
The Index measures Gemini behaviour as one constituent of the same signals that drive AI Overview measurement, with weighting calibrated to Gemini's particular surface preference. Subjects engaged on a Gemini-led brief are reviewed quarterly with a stable prompt set across both consumer Gemini and the Workspace surfaces.
A closing observation
Gemini is the system that makes the broader argument concrete: AI discovery is not, in the end, a parallel discipline to SEO. It is the discipline of becoming a credible entity inside graphs whose definitions of credibility have shifted.
The subjects that win on Gemini are the subjects that have done the unfashionable work of becoming canonical inside Google. They are also, almost invariably, the subjects that win on the AI Overview.