What is AI training data?
Vast amounts of text, images, and content used to train LLMs and AI systems.
In brief
- Knowledge base of LLMs
- Includes web, books, articles, code
- Influences what AI "knows"
AI training data is the vast amounts of text, images, and content used to train Large Language Models and AI systems.
Being present in training data influences the probability of being cited by AI models. This is why having content on authoritative sources is important for GEO.
Frequently asked questions
Where does AI training data come from?
Mainly from public web pages collected at scale, books, articles, code and licensed data. Vendors do not publish a detailed list of their sources.
Does training data explain the answers on its own?
No. It has a cutoff date, and most assistants now supplement their knowledge with a web search at question time. For a brand, the pages read live often weigh more than what the model learned.
How do you know whether your site was collected?
You cannot check directly. Your server logs, however, show visits from AI crawlers (GPTBot, ClaudeBot, PerplexityBot and others), and your robots.txt states what you allow them to do.