Google LLM patent: SEO for entity understanding

A 2023 Google patent describes how AI systems could build an understanding of businesses, brands, products, and other entities from websites and public data. The filing outlines a process for extracting information, identifying relationships, and synthesizing what Google calls a "deep, holistic characterization" of an entity. If systems like this become more influential in search, SEO may increasingly involve helping Google understand the entity behind your content—not just the content itself.

Google has spent more than two decades helping users find information published on webpages. Whether through traditional search results, featured snippets, or AI-generated answers, the process has generally started with understanding documents. As search products become more conversational and recommendation-driven, understanding individual documents may no longer be enough.

From documents to entities

Before an AI system can recommend a business, compare products, or explain a brand, it must understand the entity behind the content. That is where Google's "Data extraction using LLMs" patent (WO2025063948A1) comes in. At first glance it looks like another content extraction system—but Google describes a broader goal: artificial intelligence should generate and enhance a deep, holistic characterization of an entity.

Google defines entities broadly: people, companies, places, objects, and concepts. Rather than simply indexing facts, the system interprets information, identifies relationships, generates summaries, and develops an understanding of the entity represented.

Four steps to entity understanding

At a high level, the patent collects information from multiple sources, interprets it, and synthesizes entity understanding. Step one identifies a domain and associated entity; webpages are processed with an LLM. Step two produces a characterization—an interpretation, not a verbatim copy of extracted content.

Attributes, relationships, and third-party data

In step three, the AI extracts presence, age, principles, services, reputation, social media sentiment, and relationships between organizational elements. Step four supplements website data with maps, job listing, business, and other third-party information. The goal is a more complete picture than any single page can provide.

Summaries, graphs, and models

The system generates entity summaries that describe brand identity, positioning, and values rather than individual pages. Hierarchical graphs organize attributes into parent and leaf nodes: services connect to audiences, locations, and differentiators. Products link to features, categories, and use cases.

The question shifts from "What information appears on this website?" to "What do we understand about this business?" The patent can include presence, reputation, and external signals such as reviews and job listings—not just website attributes.

Format-independent interpretation

A recurring theme: the AI also extracts content not structured for machine consumption. Classic extractors are often limited to predefined formats; the LLM approach interprets information regardless of format and synthesizes new, purpose-built content instead of copying text.

The website remains central but is no longer the sole source of truth. It becomes one of several inputs for the entity model—critical for AI Overviews, AI Mode, and recommendation-driven surfaces.

Webpages as evidence for SEO

Patents are not product announcements, but they reveal directions of thought. Entity understanding is not new at Google—Knowledge Graph, E-E-A-T, and reputation signals have long aimed to understand sources and actors. LLMs extend that capability: websites and public information are interpreted without requiring a fixed format.

For SEO, this means a shift in perspective. Service pages establish not only keyword rankings but also what services a business offers. Case studies demonstrate expertise, team pages identify people, reviews supply reputation signals. Visibility increasingly depends on how well Google understands the entity behind a keyword—especially when AI summarizes options instead of serving ten blue links.

Enterprise, e-commerce, and local SEO

Enterprise organizations often spread information across product pages, investor relations, press releases, and recruiting—different channels, different wording. E-commerce brands must prepare product attributes, category relationships, and reviews so AI can derive recommendations for specific use cases. Local businesses benefit especially when services, service areas, and reputation signals align across website, Google Business Profile, and third-party platforms.

Structured data, consistent NAP details, and linked author profiles further support this entity model. They do not replace substantive clarity, but they make it easier for systems to map relationships between brands, people, and offerings.

Shaping brand understanding deliberately

Entity understanding emerges from the sum of many sources. Companies should check whether website, business profiles, social media, press, and job postings tell the same story—not word-for-word, but substantively aligned. Which attributes should be associated with the brand: trust, innovation, sustainability, local expertise?

• Back claims with reviews, case studies, certifications, and expert profiles.
• Link relationships between products, locations, audiences, and brands clearly.
• Audit the entity footprint: what would an AI report about the company?
• Review enterprise, e-commerce, and local SEO for consistent entity identity.

Historically, SEO optimized individual URLs. Patents like this suggest what may matter next is who stands behind the pages—and whether AI systems reliably recognize, cite, and recommend that entity.