Context Window
The amount of text an AI can consider at once. Affects how much of your content AI can process.
What is Context Window?
The context window is the maximum amount of text, measured in tokens, that a language model can take into account when producing a response. It includes everything supplied to the model for that request: the system instructions, the conversation so far, any retrieved documents or uploaded files, and the answer being generated. Anything beyond the limit is either truncated or must be summarised, and models often attend less reliably to material buried in the middle of a very long context than to material near the start or end.
Context windows have grown from a few thousand tokens to hundreds of thousands or more in recent models, which allows whole documents and long conversations to be processed at once. But answer engines still do not pass entire websites to the model. Retrieval systems select a limited number of passages from a limited number of pages, trim them, and supply only those. The effective budget for any single source is a few passages, not the whole page.
That is why page structure matters for retrieval. A passage that fully answers a question on its own, under a heading that states the question, survives the selection and trimming process intact. An answer spread across a long page, or dependent on context from earlier sections, is likely to be cut or misread when the engine lifts a fragment into the window.
Why it matters for AI search
Answer engines work within a budget of passages per answer, so a brand's page competes for a few slots rather than being read whole. Content that puts the complete answer in one self-contained section, near the top, under a descriptive heading, fits the budget; content that meanders does not. Understanding the context window also tempers expectations for llms.txt and long documents: a curated, compact statement of key facts is more likely to be used than an exhaustive one.
Related terms
Token
A unit of text AI processes (roughly 4 characters). LLM.txt file size and content length are measured in tokens.
RAG (Retrieval-Augmented Generation)
A technique that allows LLMs to pull in real-time information from external sources (like a website) before generating an answer.
LLM.txt
A proposed standard file (similar to robots.txt) that provides a Markdown-formatted map of a website for AI crawlers to consume efficiently.
LLM (Large Language Model)
A type of AI trained on vast amounts of text data to understand and generate human-like language (e.g., GPT-4, Claude 3).
Frequently asked questions
Does a larger context window mean AI reads my whole site?+
No. Even with very large windows, answer engines retrieve and trim a limited set of passages from a limited set of pages before answering. The window sets an upper bound; retrieval decides what actually gets in.
How does the context window affect how I structure a page?+
Make each section self-contained: state the question in the heading, give the complete answer in the first sentences, and include the specifics needed to support it. A passage that stands alone survives being extracted; one that depends on earlier text does not.