LLM (Large Language Model)

A large language model is the AI system underneath tools like ChatGPT, Gemini, Claude and Perplexity: a model trained on a very large body of text to predict and generate language. It is what writes the answer. It is not, by itself, a search engine — the retrieval of current web pages, when it happens, is a separate step whose results are handed to the model as context.

That separation is worth understanding because it tells a store owner where effort pays. The model has two kinds of knowledge: what it absorbed in training (fixed, months old, biased toward widely discussed sites) and what it's given at question time (live search results, a product feed). It doesn't "know" your store in any durable sense; it either remembers mentions of it or reads about it right now. You can't change the first quickly. You can change the second this afternoon, by making the pages it reads state the facts it needs. The practical test is simple: paste a product page into an LLM and ask it to describe the product for a shopper. What it misses is what every LLM-based tool will miss.