The grounding page standard is a published content specification that defines how to structure a web page so that AI systems can reliably extract, attribute, and cite information about a specific entity. Created by German SEO expert Hanns Kronenberg and released in November 2025, it provides a practical framework for making brands, organizations, products, and people legible to generative engines like ChatGPT, Perplexity, and Google Gemini. The sections below unpack where the standard came from, how it works, and what it means for your SEO and AI visibility strategy in 2026.
Where did the grounding page standard come from?
The grounding page standard was created by Hanns Kronenberg, an AI Search Advisor and SEO expert who serves as Head of SEO at Chefkoch and founded the GPT Insights analytics project. Kronenberg published the first version of the standard on November 20, 2025, through the Grounding Page Project, positioning it as an open standard initiative for machine-readable brand management. The current active version is v1.6, updated in May 2026.
The standard did not emerge in a vacuum. It responded to a measurable shift in how search infrastructure works. In May 2026, Microsoft Bing’s team published a framework explaining how indexing for AI-generated answers differs fundamentally from traditional search, validating the core premise that grounding is a distinct optimization problem with its own rules.
The Grounding Page Project describes itself as a mental framework and discipline of factual clarity, not a technical protocol like HTTP or REST. It draws on a 2026 arXiv study by Volpini et al. titled “Structured Linked Data as a Memory Layer for Agent-Orchestrated Retrieval” as architectural support, though the project’s own documentation is transparent that none of the cited studies directly evaluate grounding pages. The alignment is conceptual, not empirically validated in peer-reviewed literature specific to this format.
One important nuance: the standard is not officially endorsed by any search engine or AI platform. As the project’s own documentation acknowledges, it does not require acceptance by large language models, just as SEO was never officially accepted by Google. It is a practitioner-led specification, currently at v1.6, that gained traction in the SEO community in late 2025 and into 2026.
How does a grounding page differ from a regular landing page?
A grounding page differs from a regular landing page in its fundamental purpose and audience. A landing page is designed to sell and convert, using emotional appeals, narrative structure, and visual persuasion. A grounding page is designed to explain and provide facts, optimized for AI systems that need to extract discrete, attributable information. The unit of optimization shifts from the keyword to the entity.
The practical differences are significant. A landing page for a software product might open with a headline like “The fastest way to grow your business.” A grounding page for the same product opens with a single declarative sentence: “[Product name] is a [category] that [function].” No adjectives. One fact per sentence. Visible timestamps on information that changes over time. The spec is explicit about this discipline.
The analogy the standard itself uses is a press kit versus a product page. Both serve legitimate purposes, but they follow different rules. A press kit exists so journalists and editors can extract accurate facts quickly. A grounding page exists so AI systems can do the same. Trying to make one page serve both functions often produces a page that achieves neither marketing impact nor AI citability.
Critics including SEO veteran Kai Spriestersbach have argued that a well-maintained About page can achieve the same result as a dedicated grounding page, provided it is fact-led, unambiguous, and regularly updated. The Grounding Page Standard itself acknowledges this: the specification can be applied to an existing About page rather than requiring a separate URL. What matters is the content discipline, not the page path.
What makes a page qualify as a grounding source for AI engines?
A page qualifies as a grounding source when it contains information that AI systems can responsibly use to construct a response: facts that are fresh, attributable, specific, internally consistent, and easy to retrieve as discrete evidence. Being indexed is no longer sufficient. A page can be crawled and ranked while still being unusable for AI grounding if its content is structured for human browsing rather than machine extraction.
Microsoft’s Bing team identified five areas where grounding measurement diverges from traditional search ranking: factual fidelity at the chunk level, source attribution quality, content freshness, coverage of specific high-value facts, and internal consistency. A page with conflicting claims about the same entity fails the last test, and grounding systems must resolve those conflicts before producing an answer, unlike search engines that can surface competing results and let the user decide.
The Grounding Page Standard defines three core qualifying elements for a grounding source:
- A stable definition: A short, verifiable statement describing what the entity is, written without promotional language, one fact per sentence.
- A clear disambiguation: A statement describing what the entity is not, which prevents semantic drift when the entity name overlaps with other concepts or competitors.
- Consistent structure: The same format and logic applied throughout so AI systems can reliably extract the right information regardless of query phrasing.
A 2026 study by Volpini et al. evaluated entity-centric page architectures across more than 2,400 evaluations and found that visible, navigable entity content outperforms JSON-LD markup alone by approximately 30% in retrieval accuracy. The gains came from restructuring content into dedicated entity pages where facts are directly visible in the HTML, not from markup additions alone. Structure in the content itself matters more than schema decoration.
How do generative engines use grounding pages to answer queries?
Generative engines use grounding pages by breaking content into smaller chunks and evaluating each chunk as evidence for constructing a response. When a system like ChatGPT, Perplexity, or Google Gemini receives a query, it does not simply retrieve documents and rank them. It reconstructs facts from patterns. When facts are missing or inconsistent across sources, the model fills gaps with plausibility, which is the mechanism behind hallucinations. A grounding page aims to eliminate that ambiguity by providing a single, authoritative, citable source of truth.
The retrieval process operates in loops. A grounding system may retrieve an initial set of sources, evaluate the evidence quality, ask follow-up retrieval steps based on intermediate results, combine evidence from multiple sources, and re-evaluate when confidence is low. This is structurally different from traditional search, where a single query produces a ranked list for a human to evaluate.
Content density determines what gets cited. A 3,000-word guide full of introductory preamble can lose a citation to a 70-word definition that directly answers the prompt. Content that reads like an answer tends to win because the model needs something it can extract and attribute. This is why the grounding page format prioritizes answer-shaped sentences over narrative flow.
Microsoft’s Bing introduced a design choice called abstention: if evidence is missing, stale, or conflicting, the AI system can decline to answer rather than guess. Traditional search does not make this judgment because it presents options for a human to evaluate. For brands, this means a poorly structured or outdated page does not just rank lower. It can cause the AI to omit the brand entirely from its response.
One important limitation: not every query triggers a grounding retrieval step. If a large language model already knows a company from its training data, it may respond from parametric memory without retrieving any live page. Grounding pages are most valuable for entities that are weakly represented in training data, including local businesses, niche brands, and organizations whose names overlap with other concepts.
Which content types are best suited to the grounding page format?
The grounding page format is best suited to content about specific, named entities where factual accuracy and disambiguation matter. The Grounding Page Standard defines 17 entity classes that warrant grounding treatment: Organization, Platform, Standard, Project, Field of Knowledge, Person, Method, Product, Publication, Dataset, Tool, Event, Group or Role, Place, Service, Award, and Legal Entity. An 18th class covers Semantic Frames for abstract meaning structures.
The format is particularly valuable for entities that are weakly represented in AI training data. Local businesses, non-English brands, niche service providers, and organizations whose names could be confused with competitors or generic terms benefit most. For multi-location businesses, each location is its own entity, and a grounding page for each location serves as the central hub from which consistent NAP data radiates to Google Business Profile listings and directory entries.
Beyond dedicated grounding pages, several existing content types naturally align with the format’s principles:
- About pages: When rewritten to be fact-led, unambiguous, and regularly updated, they can serve as functional grounding pages without a separate URL.
- FAQ pages: Direct question-and-answer structure maps well to how AI systems chunk and retrieve content.
- Service and product description pages: When stripped of promotional language and structured around verifiable facts, these become strong grounding candidates.
- How-to guides and method pages: Structured, step-based content with clear attribution performs well in retrieval systems.
The Wikipedia analogy is useful here. Wikipedia is not called a grounding page, but it operates on the same principle: it defines, categorizes, disambiguates, and cites. The grounding page format is a structured evolution of that approach, applied to any entity a brand needs AI systems to understand correctly.
How do you build a grounding page that AI engines will cite?
Building a grounding page that AI engines will cite requires seven core structural elements, combined with the discipline to maintain factual accuracy over time. The format prioritizes visible content over schema markup, and factual precision over persuasive language.
Core structural elements
The Grounding Page Standard specifies the following required components:
- H1 title: The entity name only. No tagline, no descriptor.
- Single-sentence lead definition: “[Name] is a [category] that [function].” No marketing adjectives. One fact per sentence.
- Disambiguation section: A statement describing what the entity is not, preventing semantic drift.
- JSON-LD that exactly mirrors the visible HTML: Schema markup that contradicts the visible content creates a trust conflict for AI systems.
- Prominent authority signal: A footer or imprint link that connects the page to a verified organizational identity.
- Visible dateModified timestamp: Freshness is a core grounding signal. Missing or stale timestamps reduce citability.
- Human Notice: A brief UX bridge explaining to human visitors why the page is structured the way it is.
Common technical mistakes to avoid
Several implementation errors consistently undermine grounding performance. JSON-LD that contradicts the visible HTML content creates a conflict that AI systems cannot resolve. Generic schema types like WebPage instead of the specific entity type reduce precision. H2 headings without the entity name cause isolated text chunks to lose attribution when extracted. Volatile facts stated without a date stamp signal unreliability to freshness-sensitive retrieval systems.
The sameAs property in JSON-LD is critical for AI trust. Linking out to Google Business Profile, relevant social profiles, and industry directory listings, and ensuring those sources link back, gives AI systems a cross-reference network for confirming the entity. Wikidata Q-codes have become a de facto grounding standard for AI engines in 2026, with retrieval systems cross-referencing sameAs arrays against Wikidata to validate entity identity.
One finding from the Volpini et al. research is worth emphasizing: JSON-LD markup alone produces only modest improvements in retrieval accuracy. The significant gains came from restructuring content into dedicated entity pages where facts are directly visible and navigable in the HTML. Markup supports the content. It does not replace it.
Critics including Kai Spriestersbach note that AI systems generally weight third-party sources more heavily than self-reported information. What appears on Wikipedia, in press coverage, or in industry directories carries more authority for most retrieval systems than a company’s own page. A grounding page does not replace the work of PR, thought leadership, and earning citations from sources that others trust.
Does optimizing for grounding pages affect traditional Google rankings?
Optimizing for grounding pages does not directly improve traditional Google rankings. The Grounding Page Standard is explicit on this point: it improves AI visibility but is not driven by SEO mechanics. Classic Google search and AI answer generation access the same index but evaluate content using fundamentally different criteria. Classic search prioritizes click probability, brand signals, and user engagement. AI grounding prioritizes semantic similarity, information density, and factual accuracy.
Research from Ahrefs found that roughly 80% of URLs cited by ChatGPT do not rank in Google’s top 100. A page can hold the top organic position and never appear in a ChatGPT response if its content is not structured for extraction. The reverse is also true. These are parallel visibility channels with overlapping but distinct success factors.
Google has formally separated AI data use from search indexing. Google’s documentation states that Google-Extended, the AI training and grounding crawler, does not impact a site’s inclusion in Google Search and is not used as a ranking signal. Blocking it affects AI model usage only, not organic rankings.
The practical relationship between the two channels is additive rather than competitive. GEO (Generative Engine Optimization) is rooted in the same content fundamentals as SEO: authority, relevance, structured data, and factual accuracy. A grounding page built to the standard’s specification will typically also be a well-structured, authoritative page that performs competently in traditional search. The two goals reinforce each other without being identical.
Microsoft launched the AI Performance dashboard in Bing Webmaster Tools in early 2026, giving publishers page-level citation data for AI-generated answers separately from traditional ranking data. This separation of measurement reflects the reality that grounding performance and search ranking are related but distinct outcomes that require distinct tracking. For SEO professionals managing both channels, Bing’s grounding framework is the clearest public documentation of how the two systems diverge at the measurement level.
The most accurate framing is that AI SEO extends classical SEO into a new distribution channel. Where classical SEO targets navigational and commercial search, AI visibility targets representation inside generated answers: citations, entity framing, and recommendation presence. Grounding page optimization is one structured approach to earning that representation, and it works best as a complement to a strong broader SEO foundation rather than a replacement for it. Tools like the WP SEO Agent can help you manage both channels from within WordPress, tracking performance across Google and generative engines without requiring separate workflows.
