Large language model (LLM)
Short definition
A large language model is trained on vast amounts of text and can produce human-like writing and answer questions.
A large language model, LLM, is a statistical model trained on internet-scale text, it works by predicting the next word, yet the result reads as fluent and contextually appropriate. ChatGPT, Claude and Gemini are products built on top of these models.
The model does not remember facts, it recombines patterns seen during training, which is why it can be unaware of a recent event or produce a confident-sounding answer that is simply wrong.
For businesses, LLMs speed up drafting text, summarising, and answering customer questions, but the final check always stays with a person.
A large language model is often mistaken for a search engine, yet a search engine retrieves existing pages whilst a model generates new text from patterns it learned. Forgetting this difference leads to the common mistake of treating every answer as a verified fact; for anything current or business critical, the output should always be checked against a separate, reliable source.
Why it matters
Putting an LLM to the right task turns hours of writing or research into minutes. Used without knowing its limits, though, it can send a wrong fact all the way to a customer.
Illustrative example
A spa chain drafts first replies to customer questions with an LLM, but every answer that mentions price or a booking slot gets a final human check.
