llms.txt and the Grounding Page give AI systems and LLMs orientation and extend SEO into AI search. The llms.txt file guides AI agents to the most important content on a website. The Grounding Page provides verified information about companies, brands and people.
Key takeaways:
Both building blocks complement classic SEO and make a website understandable for AI systems. Each serves a different purpose:
| Criterion | llms.txt | Grounding Page |
|---|---|---|
| Purpose | Guide to selected content | Source of facts about the entity: company, brand, person or product |
| Format | Markdown file in the root directory | HTML page with structured data |
| Readers | AI agents, coding assistants, developer tools | Search engines, AI systems via the search index, and people |
| Retrieval | directly at the URL /llms.txt | via the Google and Bing index |
| Maintenance | automatic with every change | review by a responsible person, for example quarterly |
| Documented impact | no measurable effect on AI citations | case studies showing gains, without a control group |
Both draw on the same information: services, core data and descriptions. A shared data source keeps them consistent.
The llms.txt is a Markdown file in the root directory of a website, for example https://digitalagenten.com/llms.txt. Jeremy Howard proposed the standard in September 2024 on llmstxt.org.
The llms.txt file helps large language models (LLMs) quickly find the most important content at inference time. It is intended for retrieval by AI agents, not for training.
The robots.txt controls access, and the sitemap.xml lists all URLs for search engines. The llms.txt shows AI systems selected URLs with context.
H1 as the only required element: the name of the website or companyOnly an H1 with the name is required. All other elements are recommended.
The name should match exactly the name used in the business profile, structured data and directories. This lets AI systems attribute all information unambiguously to the same entity.
Summary as a blockquote: the company in two sentencesA blockquote with “>” summarizes the offering, location and focus areas. Factual statements give AI systems more context than promotional copy.
A short paragraph below it can link to the company’s facts page.
Link lists under H2: the most important URLs with a short descriptionEach section groups links in the format “[Name](URL): Description”. Typical sections are company, services, guides and documentation.
The description explains in exactly one sentence what information the linked page contains.
Optional section: secondary content for limited context windowsAI systems can skip links under “## Optional” when context is limited. Typical optional items are the blog overview, sitemap and archives.
llms-full.txt: all content in a single fileThe llms-full.txt contains the full text of the linked pages in Markdown. Anthropic, the company behind Claude, introduced the format in 2024 together with Mintlify for developer documentation.
For company websites, the llms.txt is usually sufficient. For half of the domains that offer both versions, the file sizes differ by no more than 5% (SEOmator).
The file is already widespread: 8.6% of the 100,000 largest domains according to Cloudflare Radar have an llms.txt (SEOmator, September 2026). Log file studies show how often AI crawlers actually fetch it:
A case study by Wix (December 2025) shows more requests: with a rel=alternate link in the HTML head, OAI-SearchBot fetched the llms.txt more than 5,000 times across seven websites.
Adding such a link to the head is therefore a simple way to test this exact effect on your own website.
For classic SEO and rankings, the llms.txt is therefore neutral. Its value lies with AI agents and developer tools that retrieve content in a targeted way.
A grounding page is a page on your own domain that consolidates the key information about an entity: company, brand, person or product. Uberall describes it as a “single, authoritative, citable source of truth”.
Large language models such as ChatGPT, Claude and Gemini find the page via the search index. It answers typical questions about the offering, services and contacts, precisely and with sources. This enables AI systems to cite correctly.
Grounding pages extend the classic About Us page with verifiable information. The term GroundingPage is a registered trademark of Hanns Kronenberg.
Grounding and Grounding Page: retrieval process and sourceGrounding refers to a language model (LLM) retrieving current web content, for example Grounding with Google Search in Gemini or Grounding with Bing Search.
The facts page is the source that should be found during this process. It becomes visible as soon as Google and Bing index it.
Fact table: subject, description, evidence and ownershipThe Grounding Page Standard by Hanns Kronenberg (version 1.7, October 2026) relies on a stable definition, a clear delineation and a consistent structure.
A proven approach is a fact table with subject, description, canonical reference, scope, evidence and maintenance ownership. Related entities such as locations, products and people get their own sections.
Grounding Page example: the digitalagenten facts pageThe digitalagenten facts page consolidates core data, service areas with jump links, contacts and frequently asked questions in German and English.
An Organization schema with a service catalog mirrors the visible information. The llms.txt links to this page in its very first paragraph.
The grounding page delivers its value as a single, consistent reference. Profiles, directories and press releases adopt the same data.
This creates consistent mentions that AI systems cite in their answers.
For Generative Engine Optimization (GEO), the combination is the most robust approach: the file makes content accessible, the page secures unambiguous information.
The digitalagenten llms.txt is generated automatically from WordPress. Five steps lead to a file that maintains itself:
Tools make getting started easier: the generator in Yoast SEO, plugins for WordPress and Shopify, or MCP servers for documentation.
Yoast generates the file weekly from the five most recently updated pieces of content. For websites with many service pages, a custom selection offers more control.
Your own grounding page comes together step by step:
Each step increases the likelihood that AI answers describe and cite brands correctly. Bing recommends phrasing content directly, keeping entity names consistent and placing key information at the top.
Clean files and consistent information secure the impact of both building blocks. Common mistakes can be checked before going live:
Generative Engine Optimization (GEO) extends classic SEO with content, mentions and measurement in AI answers. Within it, llms.txt and grounding pages are low-effort technical foundations.
The greatest impact comes from combining them with mentions, directories and citable content:
The digitalagenten GEO agency implements all building blocks in one program, from the grounding page to monitoring. This keeps brands visible in AI answers over the long term.
With automated maintenance, the llms.txt requires hardly any effort and makes content accessible to AI agents.
The grounding page provides the verified facts that profiles, directories and AI answers draw on.
The greatest AI visibility comes when both building blocks are fed from one data source and are followed by mentions outside your own website.
The llms.txt sits in the root directory, i.e. at https://domain.de/llms.txt. According to Jeremy Howard’s proposal from September 2024, it is a UTF-8 text file with Markdown formatting.
According to its own AI optimization guide (as of July 2026), Google ignores the llms.txt. The file affects rankings in Google Search neither positively nor negatively.
The llms.txt links to the most important pages with a short description. The llms-full.txt contains the full text of these pages and is best suited to developer documentation.
A grounding page is created in four steps: collect information with evidence, define a structure with a definition and fact table, add an Organization schema and review it regularly.
The About Us page tells the story and values. The grounding page consolidates verifiable information in a fixed structure, phrased neutrally and backed by evidence. The two pages complement each other.
The llms.txt links to the facts page and the core service pages. Both are maintained from the same data source and therefore stay unambiguous and consistent.
Lorenz Wacker is Managing Partner and Digital Strategist at digitalagenten in Berlin and specializes in comprehensive digital strategy.
As managing director, he combines SEO, ads, content and AI into multichannel strategies.
Lorenz has industry experience in automotive, financial services, insurance, online media, telecommunications, retail, travel and healthcare.
You are currently viewing a placeholder content from Vimeo. To access the actual content, click the button below. Please note that doing so will share data with third-party providers.
More InformationYou are currently viewing a placeholder content from YouTube. To access the actual content, click the button below. Please note that doing so will share data with third-party providers.
More InformationYou need to load content from reCAPTCHA to submit the form. Please note that doing so will share data with third-party providers.
More InformationYou are currently viewing a placeholder content from Facebook. To access the actual content, click the button below. Please note that doing so will share data with third-party providers.
More InformationYou are currently viewing a placeholder content from Instagram. To access the actual content, click the button below. Please note that doing so will share data with third-party providers.
More Information
