A large language model, or LLM, is the kind of artificial intelligence behind tools like ChatGPT, Claude, and Google’s Gemini: software trained on enormous amounts of written text so it can predict likely words and produce fluent, human-sounding language. When you ask one a question, it is not looking up a single stored answer, it is assembling a reply from patterns it learned across billions of sentences.
For a small business, the reason this term is worth knowing is simple. These models increasingly sit between your customer and your website. Someone asks an assistant for a recommendation, and the model answers in plain prose, often naming specific businesses. What it knows about you came from text it read: your own pages, plus what others have written.
Two things follow. First, a model can be confidently wrong, so the facts it states about your business are only as good as the sources it drew from. Second, if your website is hard for an AI crawler to read, you are underrepresented in what these models can say about you at all.
I build sites that these models can reach and read as clean text, because an assistant can only describe and cite a business whose pages it can actually take in. See AI search for how this plays out when a customer is looking for you.