Context window
The context window bounds the information a model can process in a request. Instructions, conversation history, supplied material and retrieved content can consume the available budget; input and output limits depend on the model.
How the concept fits together
- Step 1Instructions and history
- Step 2Question and retrieved material
- Step 3Available context budget
- Step 4Generated response
What it means for GEO
A crawlable document is not necessarily included in an answer’s context. State important facts and limitations together so useful excerpts remain understandable.
What not to infer
A large advertised window does not prove that the assistant read your entire website or weighted every passage equally.
How it works
A context window is the space available for information processed in a model request, not a promise to ingest a whole website. Instructions, prior messages, tool results and other inputs compete for that space, with model-specific input and output limits. Search products also decide which sources to fetch and which material to include. A published model limit tells you neither how many pages the product searched nor how faithfully it will use every condition contained in a document.
GEO example
Illustrative scenario, not observed assistant behavior or measured results.
Illustrative scenario: an integration guide says an export is available, but its account-tier restriction appears much further down the page. An assistant answering “Can a small team export audit results?” might encounter only the feature passage and omit that restriction. Rewrite the passage to state the feature and the eligible tier together, then retain the full details below. This improves the passage’s completeness without claiming that the assistant will retrieve it or that a larger context window solves selection errors.
What you can do
- Make important sections understandable without requiring the introduction or a distant footnote.
- Place exceptions next to the claim they qualify, especially pricing, availability and technical limits.
- In your own application, inspect the context actually supplied to the model and document any truncation.
What to measure, and what it cannot prove
Check whether answers preserve the specific conditions you published. Record omissions separately from incorrect facts and from absence of a citation. If you do not control the assistant, an omission does not establish that its context window was full; retrieval, selection and generation can also explain it.
Evidence and authoritative sources
- Google AI for Developers: Understand and count tokens
Documents context windows and model-specific token limits, not the source-selection behavior of every consumer assistant.
- OpenAI: Searching the web with ChatGPT
Describes web search and warns that results and citations may be incomplete or incorrect.
The sources establish the documented concepts and product behavior. The publishing checklists, scenarios and operational measurement rules are GeckoCheck educational interpretations, not provider-endorsed ranking instructions.
Source: Google AI for Developers: Understand and count tokens · OpenAI: Searching the web with ChatGPT
Period: Reviewed October 12, 2026. Living provider documentation; cited research published in 2023 and 2024 where applicable.
Sample: Conceptual explanation and illustrative scenario. No measured observations.
Limitation: GeckoCheck educational interpretation. Examples are fictional; formulas use declared operational definitions, not universal provider metrics. Source documentation does not guarantee citation or ranking outcomes.