LLM(large language model)
Also called: language model
Definition
An LLM, short for Large Language Model, is an artificial intelligence model trained on very large amounts of text: web pages, books, articles, code. It learns to predict the most likely continuation of a text, which lets it answer questions, summarize, translate or write.
An LLM doesn't look things up in a database of facts: it produces text from what it learned. That's why today's assistants combine it with web search, so they can rely on recent pages rather than on their memory alone. GPT, Claude, Gemini, Llama and Mistral are families of LLMs.
Example
When you ask an assistant to recommend software without it searching the web, it answers with what its LLM retained during training, sometimes several months out of date.
What it means for your AI visibility
What a model knows about your brand depends on what existed on the web before its cutoff date. Hence the value of having accurate, consistent information, picked up by reliable sources, well in advance.
Related terms
- RAGRAG lets an AI look for documents before answering, then write its answer from what it found.
- Knowledge cutoffThe knowledge cutoff marks the end of a model's training data. Whatever happened after it escapes the model, unless it searches the web.
- HallucinationA hallucination is false information presented confidently by an AI: an invented price, a feature that doesn't exist, a fake quote.
- LLMOLLMO is the optimization of a brand and its content for large language models, the ones that power ChatGPT, Claude or Gemini.
- PromptA prompt is the question or instruction you give an AI. In AI visibility, prompts are the questions your customers actually ask assistants.
Updated on · All glossary terms