Training Data
Definition
Training data is the text, code, and other media used to teach a machine learning model its parameters before deployment.
Full definition
For LLMs, training data often includes large web crawls, licensed datasets, and other corpora. After training, some products still fetch live pages for grounding; others answer mainly from model weights.
Publishers use robots.txt tokens (for example GPTBot or Google-Extended) to express preferences about training-related crawling where operators honor them. See AI Crawler and robots.txt.