Definition
An AI model trained on massive text data that can generate, summarize, and reason about language. GPT, Claude, and Gemini are LLMs. The core technology powering most modern conversational AI. It learns patterns in language rather than storing a database of facts, so it produces the most plausible next words rather than looking anything up.
What it looks like in practice
An electrician's website chat runs on a large language model. A homeowner types: my panel is making a buzzing sound and half the basement outlets are dead; is that urgent? Nobody wrote a rule for that exact wording, but the model has seen enough language to recognize a possible fault, reply in plain English that buzzing plus dead circuits should be looked at today, and ask for the address and a callback number. The same model, given different instructions, drafts the follow-up text an hour later. It is one general engine doing several jobs, not a script.
Why it matters
LLMs are the engine behind nearly every AI tool a small business touches: the chat widget on a site, the assistant that drafts an email, the system that answers a call. Because one model can be pointed at many jobs, an owner almost never buys an LLM directly; they buy a service built on one and inherit both its fluency and its quirks. The important quirk: these models predict plausible language, they do not look up verified facts. That is why the useful question is not which model a vendor uses, but what the model is allowed to say without a person checking it first.