Risk 01
Hallucinations
The system fills missing facts with plausible but incorrect information.
STANDARD V1.5 STABLE
Open standard for stable, machine-readable facts in AI systems.
Grounding Pages help organizations define their most important entities, brands, and concepts so AI systems can interpret them consistently, classify them correctly, and cite them more reliably.
They create a verifiable factual layer that AI systems can use for recognition, assignment, and citation.
- Stable entity interpretation
- Machine-readable facts
- Better AI citation readiness
Which risks Grounding Pages reduce
When clear, verifiable facts are missing, AI systems may misinterpret, mix up, or ignore entities. Grounding Pages address four structural risks.
Risk 02
Entity confusion
The system mixes up similar names, categories, competitors, or generic concepts.
Risk 03
Non-mention
The entity is not considered in relevant answers because the signals are not strong, clear, or trustworthy enough.
Risk 04
English-dominated retrieval patterns
Local or non-English entities may be disadvantaged because AI retrieval often favors English-language sources and English-shaped source patterns.
Grounding Pages create a stable factual layer for this: clear entity definitions, structured statements, citable evidence, and machine-readable signals.
Mission
Designed for Brand Managers and AI-SEOs, this standard establishes a framework for machine-readable brand management in the era of artificial intelligence.
It equips organizations with a precise tool to define their relevant brands, entities, and frames, directly addressing the structural risks of modern AI:
AI systems rely on pattern reconstruction, which can lead to hallucinations, missing facts, and unstable entity interpretation. The Grounding Page Project addresses this by targeting visibility deficits for weakly represented entities and English-dominant retrieval biases.
Grounding Pages provide a stable foundation of machine-readable facts that AI systems can use for interpretation. This supports semantic stability, answer quality, and entity accuracy.
The goal is to give organizations a structured method to define their core entities for AI systems.
The current evolution of search infrastructure confirms this direction: for AI-generated answers, what matters is not only which page is found, but which facts can support an answer.
Scope
This standard defines clear boundaries to support reliable interpretation by AI models.
The standard works across different entity types: organizations, people, products, services, features, concepts, methods, events, and other defined classes of the Grounding Page ontology.
In Scope
- Grounding for RAG systems
- Entity-level grounding
- Stable factual definitions
- Citation-ready structure
- Disambiguation rules
Not in Scope
- Marketing claims
- Subjective interpretations
- Dynamic or live data
- Regulated advice
- SEO manipulation
Is this extra work?
The standard does not require separate infrastructure. It is a mental framework. Like a style guide for facts, implementation can happen on existing pages, the About page, or dedicated page types. What matters is a consistent, verifiable structure. Wikipedia articles follow the same principle. And they are among the most successful content on the web.
Common concerns answered →
The standard does not require separate infrastructure. It is a mental framework. Like a style guide for facts, implementation can happen on existing pages, the About page, or dedicated page types. What matters is a consistent, verifiable structure. Wikipedia articles follow the same principle. And they are among the most successful content on the web.
Common concerns answered →
Why This Standard Exists
AI systems tend to perform better when they receive structured, consistent information. Grounding Pages create a stable semantic anchor that can support interpretation in ChatGPT, Google AI Search, Perplexity and other LLMs.
The standard is designed for RAG systems and grounding APIs (Gemini, Perplexity, Claude, Qwen, etc.).
The Bing Blog describes grounding as a different quality problem from traditional ranking: what matters is not only whether content can be found, but whether the underlying evidence is accurate, fresh, attributable, and consistent enough to support an answer. Grounding Pages address this requirement at the entity level.
The Paradigm Shift
Classic SEO optimizes documents for keywords and rankings. AI SEO is fundamentally different: it is about curating entities for stable, probable, and correct mentions in AI-generated answers. Without hallucinations.
This shift requires a new mental model. It is no longer about which page ranks on position 1, but about whether AI systems understand what an entity is, what it does, and how it differs from others. The unit of optimization changes from the keyword to the entity.
External validation: Microsoft describes the same system shift
In May 2026, Microsoft described the shift from classic search indexing to grounding indexing. While classic search optimizes for which pages users should visit, AI systems increasingly need to identify which information can responsibly support an answer.
This shifts the unit of optimization from ranking pages to verifiable, citable facts with clear provenance.
This is exactly where the Grounding Page Standard applies.
Reference: Microsoft Bing Blog, "Evolving role of the index: From ranking pages to supporting answers", May 6, 2026
This project exists to support this task. It provides the mental framework, the structural guidance, and the implementation tools. Without prescribing that everything must be implemented exactly one way, the goal is awareness and enablement, not rigid compliance.
Research
Research Signals
The Grounding Page Standard is not based on a single study. It draws on a growing body of research that points in the same direction: AI retrieval systems perform better when web content is structured around clearly defined entities.
Entity-Centric Pages for RAG Pipelines (2026)
A 2026 paper titled "Structured Linked Data as a Memory Layer for Agent-Orchestrated Retrieval" evaluates different document formats for RAG pipelines across more than 2,400 evaluations. The study shows that JSON-LD markup alone provides only marginal improvements. The largest gains occur when entity facts, properties and relationships are clearly materialized and navigable within the page content itself. The authors describe this architecture as "enhanced entity pages", built around the principle of one page per entity.
This direction closely aligns with the architectural idea behind Grounding Pages.
Volpini, Andrea et al. (2026). Structured Linked Data as a Memory Layer for Agent-Orchestrated Retrieval.
Paper on arXiv
Generative Engine Optimization (2025)
Research by the University of Toronto on Generative Engine Optimization (GEO) suggests that brands can benefit from treating their websites as an API for AI systems. While AI engines tend to favor earned media, a structured technical foundation is identified as a relevant factor for visibility. The study also highlights strong differences in source diversity across AI engines, suggesting that a single, centralized source of truth is more practical than engine-specific optimization.
Grounding Pages reflect a comparable approach by providing a reusable standard for brand-owned data.
Generative Engine Optimization: How to Dominate AI Search (Sep. 2025).
Paper on arXiv
Identity Pages and AI Interpretation (2025)
An independent analysis of more than 17,000 URLs (cross-domain, mixed entity types) found patterns suggesting that AI systems frequently draw on clear identity pages such as About pages when interpreting brands. Grounding Pages represent a structured evolution of the classical About page, providing a machine-readable facts layer for modern AI retrieval.
Microsoft Bing Blog: From ranking pages to supporting answers (May 2026)
Microsoft AI describes the shift from traditional search indexing to grounding indexing and explains why AI systems need reliable, citable information instead of only ranking pages.
Microsoft Bing Blog (2026-05-06). "Evolving role of the index: From ranking pages to supporting answers"
How to Use
Implementation follows three steps:
- Identify entities that require stable definitions.
- Create the page using the structure defined in the specification.
- Add a footer link similar to an About page or an imprint to support discoverability and structural consistency.
New · Start here
Grounding Playbook
Understand the 10 goals behind a Grounding Page before implementing the technical standard.
Open the Playbook
Free tools · Diagnose & deep analysis
Grounding Check & Entity Decoder
Paste a URL or text into the Grounding Check for a fast diagnosis, or run the Entity Decoder for a deep, multi-pass entity-centred analysis with grounding readiness scoring.
Open the Tools
Results · daily
AI Model Knowledge Comparison
Daily test of what current AI models know without live web search. Shows concretely why grounding remains essential for current facts.
View latest resultsThe Standard
The Grounding Page Standard defines structure and technical rules for AI-optimized factual pages.
Examples
Example implementations illustrate how Grounding Pages can support semantic stability and precise entity definitions.
Reference Examples
Standard-compliant examples created by the Grounding Page Project.
Implementations
Real-world adoptions on external websites.
58 verified examples listed (Updated: March 29, 2026)
About
The Grounding Page Project is an independent, open initiative. It maintains a freely available standard that aims to support answer consistency across AI systems.
Concept and architecture by Hanns Kronenberg