AI Search

What Is Large Language Model (LLM)?

Definition

A deep-learning model trained on massive text datasets to generate human-like text, power chatbots, answer questions, and summarise content. Examples include GPT-4/5, Claude, and Gemini.

Why It Matters

LLMs are the engines behind ChatGPT, Claude, Perplexity, and AI Overviews — the tools reshaping how people find information. Understanding how LLMs work helps marketers optimise content so it gets retrieved and cited when an LLM generates an answer.

How It Works

LLMs are trained on billions of tokens of text, learning statistical patterns of language. When given a prompt, they predict the next token based on patterns learned during training. Modern LLMs combine this base capability with retrieval-augmented generation (pulling live web content into the prompt context), tool use, and fine-tuning for specific tasks.

Real-World Example

When a user asks Claude "what's the average SEO cost in Australia", Claude retrieves recent articles (if given retrieval tools) or generates an answer from its training data. Content that's clear, specific, and well-indexed is more likely to be cited in either mode.

Quick Facts

GPT, Claude, Gemini, and Llama are the most widely used LLM families
LLMs don't "know" facts — they predict likely text based on training patterns
Retrieval-augmented generation (RAG) grounds LLM answers in live sources
Hallucination — confidently generating false information — is an unsolved LLM problem