Large language model

A large language model (LLM) is an AI model with a very large number of parameters that generates text by predicting what is likely to come next. Google's machine learning glossary defines it as, at a minimum, "a language model having a very high number of parameters," and more informally as any Transformer-based language model, such as Gemini or GPT. A language model, in Google's words, "estimates the probability of a token or sequence of tokens occurring within a longer sequence of tokens." Anthropic describes Claude as a family of large language models.

Why it matters for brands

An LLM learns from training data that stops at a cutoff. Anthropic, for example, publishes a reliable knowledge cutoff for each Claude model: the date through which its knowledge is most extensive and reliable. A model answering from memory alone can repeat an old price or miss a product you launched since. AI search engines fetch current pages and generate the answer from them, an approach called retrieval-augmented generation that grounds the answer in sources it can cite. Models can also state false things as if they were fact, which is called hallucination.

How it shows up in measurement

The same model can answer the same prompt differently each time. In a 2026 SparkToro study, ChatGPT, Claude and Google's AI features returned the same list of brands less than once in 100 runs, so measure across many runs, not one. Record the mode as well. In Claude, for example, turning on web search gives the model current pages to work from, and every response then includes citations. Answers with and without search draw on different material, so keep them in separate trends.