Grounding
What is grounding?
Grounding is the practice of connecting an AI model’s answer to specific, verifiable sources, such as search results, documents or databases, so that the answer reflects those sources rather than only the model’s training. A grounded answer can be checked against the pages it cites. Google, for example, offers “Grounding with Google Search” for developers using its Gemini models.
Grounding is usually done with retrieval-augmented generation: the system retrieves relevant pages and instructs the model to answer from them. It reduces, but doesn’t eliminate, made-up answers.
Google’s Google-Extended control in robots.txt covers grounding. Disallowing Google-Extended tells Google not to use your content for training Gemini models or for grounding in the Gemini apps and Grounding with Google Search for developers. Google says it doesn’t affect inclusion or ranking in Google Search.
Why grounding matters for SEO
Grounded answers are where AI citations come from. When an assistant grounds an answer in your page, your site appears as a source, and the facts in the answer come from you.
It also makes your content’s accuracy matter more. If an assistant grounds an answer in an outdated page of yours, it repeats the outdated facts with your name attached.
How to control grounding on your content
- To opt out of Gemini training and grounding outside Google Search, add a Google-Extended group with Disallow: / to your robots.txt. Google Search, AI Overviews and AI Mode are not affected.
- To limit what Google’s own AI features quote from a page, use nosnippet, data-nosnippet or max-snippet, knowing they also limit normal search snippets.
- For other assistants, allow or block their search crawlers, such as OAI-SearchBot, Claude-SearchBot or PerplexityBot. Blocking a training crawler alone does not stop grounding in live answers.
- If you want to be used as a source, keep key facts accurate, dated and stated plainly in the page text, where a large language model can find them.
Grounding example
User-agent: Google-Extended
Disallow: /Grounding FAQ
What does Google-Extended block?
Disallowing Google-Extended stops Google using your content to train Gemini models and to ground answers in the Gemini apps and Grounding with Google Search. It does not remove you from Google Search, AI Overviews or AI Mode.
Does grounding stop AI hallucinations?
It reduces them but doesn’t stop them. A model can still misread or misrepresent a source, so cited answers are worth checking.
Related SEO terms
Retrieval-augmented generation (RAG)
Retrieval-augmented generation (RAG) is when an AI system fetches relevant documents first and then writes its answer from them.
AI citations are the links an AI assistant shows as sources for its answer. Being cited means your page helped build the answer.
A large language model (LLM) is an AI model trained on huge amounts of text to predict and generate language, such as GPT, Gemini or Claude.
AI crawlers (GPTBot, ClaudeBot)
AI crawlers are bots that AI companies use to collect web content for model training or for AI search answers, such as GPTBot and ClaudeBot.