What is a large language model (LLM)?

By the Variask team · Updated

Definition

A large language model (LLM) is an AI model trained on very large amounts of text to understand and write language. LLMs power assistants such as ChatGPT, Gemini, Claude, Grok and DeepSeek.

Also called LLM, Foundation model

With Variask: Variask asks the questions your buyers ask across 9 AI engines, several times each, and reports visibility with a likely range so you act only on real change.

How LLMs work

An LLM learns patterns in text during training, then writes by predicting the next piece of text, one token at a time. What it knows comes from its training data up to a knowledge cutoff. Assistants built on LLMs add tools such as web search, so they can work with fresh information.

LLMs and brands

  • What an LLM says about you comes from what was written about you across the web, in training and in live search.
  • Different LLMs were trained on different data, so they describe brands differently.
  • Answers change with wording, language, location and from one ask to the next.

What it means for brands

  • Track several assistants, because each LLM has its own view of your market.
  • Ask the same question more than once; single answers are noisy.

Questions

Large language model (LLM):
common questions

ChatGPT is an assistant app built on OpenAI's GPT family of LLMs, plus tools such as web search.

That depends on the provider and your settings. It does not instantly change what the model tells other users.

Related terms

Keep exploring

How AI search works

Knowledge cutoff

A knowledge cutoff is the date after which an AI model has no information from its training data. Anything that happened later is unknown to the model unless it searches the web or is given the facts.

Read

How AI search works

Tokenization

Tokenization is how an AI model splits text into small pieces called tokens, often parts of words, before it reads or writes. Limits, costs and context windows are all measured in tokens.

Read

How AI search works

Context window

A context window is the amount of text an AI model can take in at once, including the question, the conversation so far and any documents or search results added to it, measured in tokens.

Read

How AI search works

AI hallucination

An AI hallucination is a confident statement from an AI model that is false or not supported by any source, such as a wrong price, a feature that does not exist or an invented quote.

Read

Engines and surfaces

Answer engine

An answer engine is a search tool that replies with a direct, written answer instead of a list of links. ChatGPT, Perplexity, Gemini, Claude, Copilot and Google AI Overviews all work as answer engines.

Read
Full glossary