Como Gemini AI decide que negocios locales recomendar

Gemini is not a search engine with a local results tab. It is a reasoning model that assembles responses from a structured internal understanding of the world, an understanding that is heavily shaped by Google's Knowledge Graph, Google Business Profile data, and the schema markup on individual business websites. When Gemini recommends a local business, it is the endpoint of a multi-step pipeline that most marketing teams have never mapped. Understanding that pipeline is the most direct route to improving your business's appearance in Gemini responses.

This guide walks through exactly how Gemini resolves local business queries, which signals carry the most weight, and the specific optimisation steps that move businesses from absent to consistently cited, with particular attention to the location-data layer that most SEO tools miss entirely.

The Gemini Local Recommendation Pipeline

When a user submits a query like "best physiotherapy clinic near Ghent open Saturday mornings," Gemini executes a sequence of steps that resembles a database lookup more than a web search:

Step 1: Query Decomposition

Gemini parses the query into structured components:

• Entity type: physiotherapy clinic (maps to MedicalBusiness / HealthAndBeautyBusiness schema types)
• Geographic anchor: near Ghent
• Temporal constraint: open Saturday mornings
• Implicit signals: "best" implies rating/review quality threshold

Step 2: Knowledge Graph Entity Lookup

Gemini queries Google's Knowledge Graph for entities matching the decomposed query. The Knowledge Graph is a structured database of named entities, businesses, people, places, concepts, with attributes and relationships. Your business needs to be a resolved entity in the Knowledge Graph to appear in this step.

Knowledge Graph entity strength is determined by:

• Data from your Google Business Profile (GBP)
• Structured data (JSON-LD) on your website
• Consistent citations in authoritative third-party sources (directories, local news, official tourism sites)
• Historical search engagement signals from Google Search

Step 3: Constraint Filtering

Gemini filters the candidate entities against the structured constraints from the query. This is where openingHoursSpecification in your schema does its work. A business without machine-readable opening hours cannot be confirmed as "open Saturday mornings", Gemini either omits it or flags uncertainty.

Step 4: Confidence Scoring

Each surviving entity receives a confidence score based on the quality and consistency of its data. High-confidence entities get cited. Low-confidence entities are omitted, even if they would factually be a good answer.

Step 5: Response Assembly

Gemini assembles the response, citing high-confidence entities and sometimes explaining why they match the query ("open Saturdays," "rated 4.8 by 240 patients").

[Image: Flow diagram showing: Query → Decompose (entity type, location, time, intent) → Knowledge Graph lookup → Constraint filter (hours, location) → Confidence score → Cite or omit, with annotation showing where each optimisation step intervenes]

Google Business Profile: Still the Bedrock

GBP is not a legacy product that Gemini has moved past. It is one of Gemini's primary data feeds for local entities. Every field you complete in GBP is a data point that strengthens your Knowledge Graph entity:

• Business category, primary category maps directly to schema @type; secondary categories add semantic richness
• Service area, for businesses that serve customers beyond their physical location
• Attributes, accessibility features, payment methods, amenities all feed entity attributes
• Opening hours, including special hours for holidays
• Photos, Gemini can cite businesses with photos more richly than those without
• Reviews, both volume and recency affect confidence scoring
• Q&A responses, structured question-answer content that directly matches common query patterns

The relationship between GBP and your website schema is one of corroboration. Gemini's confidence in your entity increases when both sources agree. A mismatch between your GBP hours and your website's openingHoursSpecification reduces confidence. An exact match increases it.

For a deeper dive into the full AEO picture, see our complete guide to Answer Engine Optimisation.

Entity Disambiguation: Why Coordinates Are a Gemini Priority

Entity disambiguation is one of the hardest problems Knowledge Graph systems face. If there are twelve businesses named "Green Garden Restaurant" across Europe, Gemini needs a reliable way to distinguish between them and match the right one to a location-specific query.

Geocoordinates are the most reliable disambiguation signal. When your LocalBusiness JSON-LD includes:

"geo": {
  "@type": "GeoCoordinates",
  "latitude": 51.0543,
  "longitude": 3.7174
}

...and those coordinates match your GBP listing, Gemini can resolve your entity with high confidence even when your name is common. Without coordinates, disambiguation relies on address strings, which are prone to formatting inconsistencies, or on inference from nearby landmark mentions, which is inherently uncertain.

The practical consequence: a business with coordinates in its schema is systematically more citable by Gemini than a business without them, independent of review quality or domain authority. This is the location-data layer that most SEO tools and guides never address, because traditional SEO doesn't require coordinates on your website.

The Structured Data Signals Gemini Weights Most

Based on the Knowledge Graph entity structure that GBP and Google Search data feeds, these are the schema fields with the highest leverage for Gemini citation:

The sameAs field deserves specific attention. Linking your schema to your GBP URL, your Yelp page, your Facebook page, and relevant industry directories creates a corroboration network, multiple authoritative sources all pointing to the same entity. Gemini uses these links to verify that its Knowledge Graph entry matches the real-world business.

"sameAs": [
  "https://www.google.com/maps/place/YOUR_BUSINESS_ID",
  "https://www.facebook.com/yourbusiness",
  "https://www.yelp.com/biz/your-business"
]

For the complete JSON-LD implementation, see our JSON-LD schema guide for local businesses.

NAP Consistency as a Confidence Multiplier

Name, Address, and Phone number consistency is a foundational signal for Knowledge Graph confidence. Every time Gemini's entity resolution process encounters a source where your NAP matches your website and GBP exactly, the confidence score for that entity increases. Every time it encounters a discrepancy, a slightly different street abbreviation, a local phone number versus a national number, a trading name versus a registered company name, the confidence score decreases.

The compounding nature of this is important: a business with perfect NAP consistency across 15 sources is not just 15x more likely to be cited than one with consistency across 1 source. The confidence score improvement is non-linear, because multiple corroborating sources resolve ambiguity that any single source cannot. Read more in our guide to NAP consistency for AI search.

[Image: Side-by-side comparison of a Google Business Profile listing and a website JSON-LD code block, with matching fields highlighted in green: business name, address, geocoordinates, opening hours, demonstrating the corroboration signals Gemini weights most]

The Review Content Signal

Gemini does not only count reviews, it reads them. Review content that contains location-specific language ("great spot in the Jordaan," "easy parking on the corner of Rue de Rivoli") reinforces the geographic entity association. Reviews mentioning specific services or specialisms reinforce category and attribute associations.

This means a response strategy for reviews is not just about volume. Encouraging reviewers to mention specific aspects of their experience, location, specific service, specific staff interaction, generates richer entity attributes than generic positive reviews.

Practical Optimisation Steps

Bring the pipeline understanding into concrete action:

1. Complete your GBP fully, every field, every category, current hours including special hours, at least 10 photos
2. Implement LocalBusiness JSON-LD with geo, openingHoursSpecification, priceRange, areaServed, and sameAs
3. Verify coordinate consistency, cross-check your schema coordinates against your GBP pin location
4. Build your sameAs network, ensure consistent listings on Yelp, Facebook, Apple Maps, and relevant industry directories
5. Fix NAP inconsistencies, audit every major platform and standardise formatting
6. Maintain review recency, implement a post-purchase or post-visit review request

Use the free MapAtlas AEO Checker to audit your current structured data signals, then review the full AI Search Visibility solution to understand how MapAtlas connects your mapping infrastructure to your Gemini citation pipeline.

The Map-Data Layer Most SEO Teams Miss

Traditional SEO tools can audit your title tags, your backlink profile, and your page speed. They cannot audit whether your geocoordinates in your schema match your GBP pin, whether your service area markup is correct, or whether your location data is consistent across mapping platforms. This is the gap MapAtlas is built to close, connecting the structured geodata that feeds Gemini's Knowledge Graph with the verification and monitoring tools businesses need to maintain it over time.

In a world where Gemini is increasingly the first point of contact between a consumer and a local business recommendation, that gap is not an SEO technicality. It is a revenue question.