Large Language Model (LLM)

An LLM is a neural network, typically a transformer, trained to predict the next token in a sequence across trillions of words of text. Through that single objective it absorbs grammar, facts, reasoning patterns, and associations between entities. Scale is the defining trait: billions to trillions of parameters encoding statistical knowledge of the written world.

The models behind ChatGPT, Claude, Gemini, and Perplexity are all LLMs, differentiated by training data, size, post-training methods, and the retrieval systems wrapped around them.

An LLM does not look facts up in a database; it generates them from learned associations. If your brand co-occurs with "best CRM for startups" across thousands of reputable pages, the model internalizes that association and reproduces it when asked for recommendations. Sparse or contradictory coverage produces hedged, generic, or wrong answers.

This is why AI answers are a lagging mirror of web reputation. The model's parametric memory rewards years of consistent presence on authoritative pages, something no quick technical fix replaces.

LLM-generated answers increasingly intermediate buying decisions: users ask for shortlists and the model names brands directly, with or without links. Influencing that output is the core of generative engine optimization, shaping both training-time presence and retrieval-time citability.

Because each lab's model learned from different data, your representation varies by engine. Geonimo queries the major LLM platforms daily with your category prompts, turning opaque model behavior into a measurable visibility score per engine.