GEO Glossary
LLM
An LLM (large language model) is a model trained on huge amounts of text that generates text by predicting the most likely next word, word by word.
At a glance
INPUT Prompt → split into tokens │MODEL Training data + optional web search │OUTPUT token by token, the most likely continuation
Example
ChatGPT, Claude, Gemini and Perplexity are built on LLMs. The model itself only "knows" what was in its training data. It only gets current information when the product around it starts a search and passes along the results.
In practice
For GEO there are two routes into the answer: what the model learned about you in training, and what it finds live in a search. You influence the first in the long term through your presence online, the second through content that ranks for the right searches.
How it fits together
What an LLM knows without search comes from its Training data. With search, RAG and Grounding come into play. Without either, the model sometimes fills gaps with a Hallucination.