Large language models, explained without the hype

How tools like ChatGPT, Claude, and Gemini produce text, why they sometimes get things wrong, and how to use them well.

Prediction, not lookup

A large language model (LLM) was trained on a huge amount of text to predict what comes next. That makes it very good at writing, summarizing, and reasoning in language, but it is not a database of facts.

Why they make things up

When a model doesn’t know, it can still produce something that sounds right. This is often called a hallucination. Ask for sources, check numbers, and give it your own documents when accuracy matters.

How to get better answers

Give context (who you are, who the audience is), show an example of what good looks like, and ask it to list its assumptions. Treat the first answer as a draft.

Choosing a model

For most everyday work, the major assistants are close enough that the best one is the one your team will actually use. Compare them on your own real tasks, not on leaderboards.

Want help applying this? See how we work or browse AI tools.