Large Language Model (LLM)

A large language model (LLM) is a neural network trained on very large amounts of text to predict the next token in a sequence. Repeated across thousands of steps, that prediction produces fluent sentences, summaries, code and answers. ChatGPT, Claude and Gemini are assistants built on LLMs.

What does an LLM know, and what does it not?

An LLM holds patterns learned from its training data, not a database of verified facts. Its knowledge stops at the knowledge cutoff, and it can state wrong information with confidence, a failure known as hallucination. Many engines reduce that risk by retrieving live sources before answering, a technique called grounding.

Why do LLMs matter for communications teams?

LLMs now sit between brands and buyers. When a buyer asks an answer engine which vendor to choose, the model composes a response from what it learned in training and what it retrieves. A brand shows up in that response through a consistent entity profile, strong entity authority and pages written so passages can be lifted cleanly. Measuring how often a brand appears is the job of Citation Share.

Different models behave differently. ChatGPT, Claude, Gemini and Perplexity weight sources in different ways, which is why cross-engine breadth is tracked rather than a single engine.