Language-based generative AIs use a Large Language Model (LLM) Generative Pre-trained Transformer (GPT). This model uses hundreds of millions of data sources, usually scraped from the Internet, to calculate, predict, and produce a response to a prompt.
GPT does not think or understand like a human. It guesses or predicts the next word in a series of words, similarly to a phone's autocorrect or the autofill in email software.