What actually is an LLM, and why do people say it just predicts the next word?

An LLM, or Large Language Model, is a type of AI trained on massive amounts of text to predict the most likely next word in a sequence, which is how it generates coherent sentences, paragraphs, and even entire articles. Saying it 'just predicts the next word' is a bit like saying a car 'just burns gasoline' — it's technically true but misses the emergent complexity. When you ask it to write a poem about a cat, it isn't searching a database. It's calculating, word by word, what should probably come next based on patterns it learned from reading a huge chunk of the internet, books, and other text. The 'large' part is key. These models have billions of parameters, which you can think of as tiny dials that get tuned during training to recognize everything from grammar rules to the emotional tone of a sentence. A smaller model might finish 'The cat sat on the...' with 'mat' because that's the most common pairing. But a large model, given the start of a sad story, might predict 'cold, hard floor' because it has also learned about narrative mood and context. This prediction process happens over and over, with each new word being fed back into the model to generate the next one. That's how it builds a full response. I've found that understanding this helps with using the tools better. If you get a weird answer, it's not because the AI is confused in a human way — it's because the statistical path it followed led to a strange place. For a deeper dive, see our guide on how to write AI prompts. **Related**: How is an LLM different from a regular chatbot? | Can an LLM actually understand what I'm saying?