Local SEO gets clearer once you can see how Google actually builds a place, and a recovered set of internal Google Maps material now maps that system out. The material exposes 72 ranking signals inside Google’s Oyster Rank system, 793 data source providers, 446 local search intent types, 50,998 Mapcore styles, 12,936 label styles, and 10,936 searchable Geostore declarations. Together they show that the Maps listing you see is the final output of an entity model, several ranking systems, geographic retrieval, personalization, and a rendering engine, not a single algorithm you can reduce to a checklist.<\/p>\n
Google represents geographic objects internally in a system called Geostore, where each object is a Feature. A Feature can be a business, building, road, city, station, area, transit element, or a 3D object. For an establishment, that object can hold identity, geometry, source information, websites, business-chain relationships, Knowledge Graph references, concepts, and ranking information.<\/p>\n
The Maps listing is assembled later from this canonical representation. What a business owner edits in Google Business Profile is not necessarily what Google maintains internally. The canonical entity can pull data from multiple sources, survive changes in geometry, and connect to other Google identifiers, including the Knowledge Graph machine ID (MID). For local SEO, the entity is the more useful unit to work with, and the listing is the interface sitting on top of it.<\/p>\n
Geostore uses a provenance system. A single business can draw its name from one provider, its phone number from another, its category from another, and its geometry from somewhere else. The corpus exposes 793 source providers along with mechanisms for provenance, priority, trust, and conflation.<\/p>\n
Conflation is the process Google uses when several sources describe the same object and disagree. Geostore can pick one value, merge several, or combine them, and it models trust levels that range from blocked or untrusted sources up to trusted and super-trusted ones. This reframes a familiar problem: editing a field in Google Business Profile does not guarantee that Google’s canonical representation instantly changes. The edit is one more piece of evidence entering a system that may already hold competing evidence, which explains why incorrect attributes, duplicate information, or reverting changes can be harder to fix than editing a listing.<\/p>\n
Geostore has its own ranking system called Oyster Rank, and the recovered material includes a complete visible enumeration of 72 signals. Named signals include Google reviews, web query volume, listing impressions, listing opens, direction requests, website clicks, chain membership, Wikipedia signals, popularity, prominence, landmark information, and road usage. Of the 72 values, 25 are explicitly marked deprecated.<\/p>\n
The recovered data gives the signal names, not their current weights. The schema shows a pipeline that extracts raw observations, normalizes them, and mixes them into the Feature’s rank, but the coefficients that would show how much each signal contributes sit outside the recovered scope. A signal such as SIGNAL_GOOGLE_REVIEWS proves reviews belong to the Oyster Rank vocabulary. It does not prove a specific weight in any given Maps search.<\/p>\n
Oyster Rank appears to describe the importance of an entity inside Geostore. A user query still passes through additional systems. Maps has to understand what the person means, identify a geographic context, generate candidates, evaluate semantic relevance, and serve a final result set. A simplified pipeline runs roughly as entity, then query understanding, then semantic matching, then candidate generation, then geography and quality, then reranking, then results.<\/p>\n
There is more than one scorer involved. A separate scorer runs entirely offline on the device, with eight signals across 13 tiers, distinct from both Oyster Rank and server-side Places ranking. Different scoring and retrieval systems operate at different stages, so treating the 72 signals as 72 ranking factors would miss most of the structure.<\/p>\n
Testing the geographic layer directly showed a dynamic footprint rather than a fixed radius. From the same origin in Paris, a dense query such as “pharmacie” produced a far smaller search area than a brand query such as “Carrefour.” The same pharmacy query expanded dramatically when run in a sparsely populated rural area, so Google appears to adapt the candidate space to both the query and the surroundings.<\/p>\n
Removing geographic weighting from the same engine, across 5,083 calls and 86,584 results, moved the median distance from 6.87 km with geography to more than 4,000 km without it. The non-geographic order stayed extremely stable, which suggests geography changes what the retrieval system considers in the first place rather than simply reordering the same list by distance. Distance remains fundamental, but “I am closer, so I should rank higher” is an incomplete model.<\/p>\n
Geostore Features can connect to the Knowledge Graph through a MID, and documents in the web index can carry MIDs too. Google has a layer called webref that associates documents with entities and stores information including topicality, confidence, geographic metadata, and document-level scores. The recovered structures describe a relative ranking signal between different documents for the same entity, along with properties such as whether a page is an author page, publisher page, or reference page.<\/p>\n
That gives a store locator or location page a wider role. Beyond ranking for a query, the document can become evidence about the underlying entity. The work becomes making it easy for Google to establish which entity the document describes, how much of it is about that entity, how confident the association should be, and whether the page is a useful reference. Web SEO and local SEO are far less separate inside Google’s infrastructure than their interfaces suggest.<\/p>\n
The semantic layer goes beyond the primary category on a listing. Google uses GConcepts, a shared conceptual vocabulary that can describe businesses, dishes, attributes, cuisines, and service modes. A “ramen” query connected with ramen restaurants, Japanese restaurants, Asian restaurants, and related concepts. Inside listings, review topics, menu dishes, and other attributes can be represented as entities rather than plain strings, which lets Google work from structured themes and precomputed signals instead of rereading thousands of reviews for every question.<\/p>\n
Some geographic intelligence lives on the phone. The recovered material includes on-device structures for visits, place candidates, frequent places, trips, home and work, mobility patterns, and user location profiles, plus a ChainAffinity object that can model affinity toward a recurring retail chain. Personalization in Maps can combine server-side knowledge of the world with a local model of the user’s own geography.<\/p>\n
Search results are only one output of Maps. The visual map cannot label thousands of relevant entities at once, and that job belongs partly to Mapcore. The recovered material includes 50,998 Mapcore styles and 12,936 label styles, and label visibility can change with zoom and other rendering conditions. A business can be eligible or highly ranked and still not appear as a visible name on the map, which matters when people measure Maps visibility with screenshots or grids. Search ranking and map visibility are separate optimization problems.<\/p>\n
Google is expanding Ask Maps and other AI-powered experiences, and much of the infrastructure needed to answer complex questions is already present: canonical place entities, semantic concepts and attributes, reviews and extracted topics, Knowledge Graph relationships, web evidence, geographic retrieval, behavioral signals, personal geographic context, listing composition, and ranking systems. Gemini adds a conversational interface over these layers. A query such as “best ramen near me” is straightforward, while a request for where six people can eat near a hotel tonight, with one vegetarian, little waiting time, and good recent service feedback, draws on what a restaurant is, what it serves, when it is open, what people say, where it sits relative to the user, and whether the evidence is reliable enough to recommend.<\/p>\n
Oyster Rank is Geostore’s internal ranking system. The recovered material includes a complete visible list of 72 signals, such as Google reviews, web query volume, listing impressions, direction requests, website clicks, chain membership, Wikipedia signals, popularity, and prominence, with 25 marked deprecated. The names are known, but the current weights were not recovered.<\/p>\n
Not necessarily. Google builds a canonical entity in Geostore from up to 793 source providers, using provenance, trust levels, and conflation to resolve disagreements. An edit enters as one more piece of evidence, so it can be outweighed by existing data, which is why some attributes revert or stay incorrect.<\/p>\n
No. Testing showed the geographic footprint changing by query and environment. Across 5,083 calls and 86,584 results, removing geographic weighting moved the median distance from 6.87 km to more than 4,000 km, and geography appears to change which candidates are considered rather than only reordering them by distance.<\/p>\n