Glossary · Technical
Context Window
The amount of text a model can hold in memory at once — the conversation, any retrieved pages, and its own output — measured in tokens.
What it means
Every model has a fixed context window, and everything it needs to consider when generating a response — the user's question, prior conversation turns, retrieved search results, system instructions — has to fit inside it. Once the window is full, older content gets dropped or summarized to make room for new input.
Why it matters for AI visibility
When a platform searches the web before answering, only the pages that fit inside the remaining context window can actually influence the answer. If a query pulls in ten sources and the window can only hold summaries of a few, most of what was retrieved never reaches the part of the model that writes the response.
See more terms in the full glossary, or read the AI Visibility 101 guide for the full picture.