Put specialist terms into clear context
The SEO and GEO glossary
Specialist terms are helpful when they make a decision easier. This glossary explains key terms from search engine optimisation, online marketing and AI search, showing what they mean for a business website.
SEO, SEA, GEO and LLMO
SEO – Search Engine Optimization
Search engine optimisation improves a website’s content, structure and technical foundations for organic search. This includes understandable pages, relevant internal links and checking technical obstacles. Read the SEO basics.
SEA – Search Engine Advertising
SEA refers to paid search advertising. Adverts and organic results have different processes and cost models. Paying for an advert does not buy a better organic position.
GEO – Generative Engine Optimization
GEO examines how businesses and content are considered in generative answers. Relevant factors include clear business information, suitable answers and documented source references. GEO explained clearly.
LLMO – Large Language Model Optimization
LLMO is used for optimisation in the context of large language models and often overlaps with GEO. The label alone does not describe a fixed scope of services. What matters is which content, systems and measurement methods are specifically meant.

Crawling, indexing and canonical
Crawling
A crawler fetches pages automatically. Whether a request is possible depends on factors including access rules, server responses and links. A completed fetch does not yet mean indexing.
Indexing
A search engine processes and includes a page in its index. The search system makes that decision. Even an indexed page does not receive a guaranteed position for a particular question.
Canonical
A canonical reference identifies the preferred URL for identical or very similar content. It should consistently match the page destination actually intended. Incorrect canonicals can make interpretation harder.
Meta description and title
The title describes the page topic, while the meta description summarises the content. Both should precisely match the page. Search systems may assemble the actual displayed result differently.
robots.txt, noindex and llms.txt
robots.txt
The robots.txt file contains rules for automated requests by particular bots. It is not access protection for confidential content. Search access and training must be distinguished according to each bot’s purpose.
Noindex
A noindex instruction asks search engines not to include a page in their index. It is not a password barrier. Changes to indexing instructions should be limited to the specific pages affected and checked afterwards.
llms.txt
llms.txt is a proposed file format intended to guide AI systems. Its effect depends on whether a system actually uses it. Google Search ignores the file; it has neither a positive nor a negative ranking effect there. It replaces neither an accessible website nor helpful content.
OAI-SearchBot and GPTBot
OAI-SearchBot is intended for OpenAI search, while GPTBot is for model training. A decision about search access is therefore not automatically a decision about training. ChatGPT-User concerns user-initiated access.
Structured data and business identity
Structured data
Machine-readable information can describe which organisation, person or page a piece of content concerns. It should agree with visible information. Markup alone guarantees neither a special search result display nor an AI recommendation.
Organisation and person
The agency, its legal operator and responsible person may involve different information. Consistent presentation prevents contradictions between brand, location and contact details.
Internal linking
Links between your own pages show relationships and lead readers to further information. Precise link text is more useful than repeated “click here”. Destinations should be accessible and answer the question indicated.
LLM, token and context window
LLM – Large Language Model
A large language model processes and generates language using learned patterns. An application can connect it with search or other tools. The generated answer must still be checked for accuracy and suitable sources.
Tokenisation
Tokenisation breaks text into processing units. A token is not necessarily a complete word. This provides no useful rule for writing website content to a particular token count.
Context window
The context window describes the amount of information a model can consider in one processing step. In practice, it therefore also matters which documents and previous messages are actually included.
How does ChatGPT work in everyday business?
Depending on the function used, the application can generate language, work with supplied content and research current information. A business should define the specific function, permitted data and necessary review effort.
RAG and fine-tuning
RAG – Retrieval-Augmented Generation
With RAG, information relevant to a task is retrieved from a source and supplied for the answer. The quality, freshness and authorisation of those sources are decisive. RAG does not prove that any arbitrary public website is automatically used.
Fine-tuning
Fine-tuning adapts a model to examples or desired behaviours through further training. It differs from current web search or editing a business website. For many editorial tasks, a good workflow with review is initially more important than adapting a model yourself.
Mention, source citation and conversion
Mention and recommendation
A mention is the appearance of a clearly attributable name. A recommendation assesses it as a suitable option for a question. Not every mention is a recommendation.
source citation
A source citation links a specific source in an answer. Check whether the linked page actually supports the statement. A link alone is not a quality seal for every statement in an answer.
Conversion
A conversion is a predefined action, such as an enquiry received. A phone click, a submitted form and an order are different events. A sound evaluation states explicitly what is counted.
The guide Measuring AI visibility connects these terms with a practical review process.
Your next step
Let’s discuss
your website.
What should your website achieve for your business? Tell us about your plans.

