Large language model (LLM)
Also known as: LLM, Language model, Foundation model
A large language model (LLM) is an AI model trained on vast amounts of text to predict the next word in a sequence, which lets it answer questions, summarize documents, write and hold conversations.
OpenAI’s GPT models behind ChatGPT, Anthropic’s Claude models and Google’s Gemini models are all large language models. They’re trained on broad collections of text such as books, web pages and code, then tuned with human feedback so they follow instructions and hold a conversation.
Where the model’s knowledge comes from
An LLM’s answer can come from two places:
- Training data: what the model learned from the text it saw up to its training cutoff. It doesn’t know what happened after that date.
- Live search: many assistants search the web before answering and use the pages they find. This approach is called RAG.
How a brand shows up in AI answers depends on both. For answers that involve search, it also matters whether the site is open to AI crawlers.
Limitations
Language models generate text based on probabilities, so they can produce fluent answers that are wrong. That’s called hallucination. They can also give different answers to the same question at different times, which is why AI visibility should be measured by asking the same questions repeatedly rather than relying on a single try.
Example
When someone asks “Which companies in Ankara make industrial molds?”, the model blends what it remembers from training with whatever pages it finds at that moment to build its list.