An AI model trained on very large volumes of text that generates language by predicting what comes next; the technology behind products like ChatGPT, Claude and Gemini.
What is an LLM?
LLM stands for Large Language Model. It is trained on very large volumes of text and fundamentally does one thing: produce the text most likely to follow the context it is given. At sufficient scale, that deceptively simple mechanism produces capabilities like summarising, translating, answering questions and writing code.
The distinction that matters for marketing: an LLM is a probability model, not a database. It does not look a fact up in records — it generates from training data and from the context it is handed. So the clearer, more consistent and more repeated the information about your brand is on the web, the more likely the model is to describe you accurately. The reverse holds too.
Why LLMs matter for brands
Customers now ask "which sunscreen suits sensitive skin" of a language model as readily as of a search engine. A brand absent from the answer simply did not exist at that decision moment. GEO is the discipline aimed squarely at this visibility surface.
- Models mostly reach current information through search and crawling, so a site closed to AI crawlers is absent from the answers too.
- Models can be wrong about your brand; the fix is not applied to the model but to the sources on the web.
- Clear definitions, FAQ blocks and consistent brand facts are the formats models quote most easily.
For the smaller, narrower sibling see SLM; for models given the ability to use tools see AI agents. The practical application is in our GEO guide.
Frequently asked questions
What is the difference between an LLM and a search engine?
A search engine ranks the pages it has crawled and hands you a list of links; you go to the source. An LLM produces an answer, and often the user never clicks anything. That changes the unit of measurement: what matters is not your position in a ranking but whether you appear — accurately — inside the answer.
What do I do if LLMs say something wrong about my brand?
You cannot edit the model, but you can edit the sources. Keeping brand facts clear, current and machine-readable on your own site, consistent across third-party sources, and keeping the site open to AI crawlers is what gradually corrects the answers being generated.
Related concepts
AI Agents
A system where a language model plans and carries out a multi-step goal by using tools; it does not just produce text — it searches, reads pages and takes actions.
AI Crawlers
The bots AI companies run to read the web for training data and live answers; GPTBot, ClaudeBot and PerplexityBot are the best-known examples.
GEO (Generative Engine Optimization)
The discipline of making sure a brand is mentioned accurately, currently and favorably inside answers generated by ChatGPT, Perplexity, Google's AI summaries and other generative engines.
RAG (Retrieval-Augmented Generation)
Having a language model write its answer from documents retrieved at question time rather than from memory alone; the method behind the citations you see in AI answers.
Search Engine
A system that crawls the web, indexes what it finds and ranks the best matches for a query; the ground on which a brand is found at the moment of intent.
SLM (Small Language Model)
A language model running on far fewer parameters than an LLM, focused on a narrow set of tasks and cheap enough to run on a single device.