A Google patent making the rounds in mid-2026 — "Data extraction using LLMs" (WO2025063948A1, filed 2023) — describes a system that reads your website and other public sources to build what Google calls a "deep, holistic characterization of a particular entity." In plain terms: an AI that figures out who your business is, what it does, and who it serves. It points to a real shift in SEO priorities — but before you rearrange your strategy, an important caveat: this is a patent filing, not a confirmed production system.

What the Patent Describes

The patent outlines a system that collects information from across the web — websites, business listings, reviews, maps data, job postings, and other public sources — and synthesizes it through AI interpretation rather than simple keyword extraction. Crucially, it's designed to process unstructured information "irrespective of its format."

What it aims to extract about an entity includes: presence, age, principles, services, reputation, social media sentiment, and the relationships between different elements of the organization. The output isn't a ranked list of pages — it's a synthesized understanding of the business itself.

The Shift: From Ranking Pages to Teaching AI Your Identity

The central idea is a meaningful reframe. Traditional SEO is about helping Google understand your webpages. This points toward helping AI understand your entity — because before an AI system can recommend, compare, or summarize a business in an AI Overview or chatbot answer, it first has to know what that business actually is.

If a model's picture of you is fuzzy, inconsistent, or contradicted across sources, you're less likely to be confidently named when someone asks a relevant question. A clear, consistent entity is easier for AI to recognize and reproduce.

Important Caveat: It's a Patent, Not a Confirmed System

Companies patent far more than they ship. This filing reveals how Google is thinking about entity understanding, but it does not confirm that this exact system is live in Search today. Treat the actions below as low-risk fundamentals that help across all of search and AI — not as a response to a confirmed ranking factor.

What To Do About It: Step-by-Step

Step 1: Audit your entity footprint

Ask the blunt question the patent implies: if an AI read everything publicly available about my business, what would it conclude? Check your website, Google Business Profile, social profiles, directories, and press for an accurate, coherent story.

Step 2: Fix consistency across sources (NAP and beyond)

Align your name, address, phone, and core description everywhere they appear. Inconsistent NAP data and conflicting descriptions across directories weaken the model's confidence in who you are.

Step 3: Strengthen your About page and structured data

Make your About page specific — credentials, history, services, and what differentiates you — instead of generic filler. Add Organization schema with sameAs links to your authoritative profiles so machines can connect the dots between your identity across platforms.

Step 4: Provide evidence and reinforce relationships

Back up claims with reviews, case studies, certifications, and press mentions. Use site architecture and structured data to make the relationships between your products, services, locations, and audiences explicit.

Quick-Win Checklist

  • Audit what a single AI view of all your public sources would conclude about your business.
  • Correct NAP inconsistencies across your site, Google Business Profile, and directories.
  • Rewrite a generic About page with specific credentials and differentiators.
  • Add Organization schema with sameAs links to your authoritative profiles.
  • Substantiate claims with reviews, case studies, certifications, and press.
  • Remember this is a patent — treat these as durable fundamentals, not a confirmed ranking factor.

Sources