Fenêtre de jetons
A Token Window (or Context Window) is the limit on the amount of text (measured in tokens) that an AI model can process in a single interaction. It determines how much of your website's content an AI can "read" and "remember" at one time to generate an answer.
Why Content Size Matters for AI Visibility
If your web pages are bloated with excessive code, redundant text, or unstructured content, they might exceed the AI's token window, causing it to "forget" the beginning of the page or ignore critical details entirely. GPT-4 has a large window (128k tokens ≈ 96,000 words), but smaller, faster models used for real-time search might only read 8k tokens (≈ 6,000 words). Critical information must be at the top, clearly structured, and concise. This is why JSON-LD schema is crucial—it presents facts in ultra-compact format that fits comfortably in any token window, ensuring AI models can always access your key data.
Human Reading vs. Token Window Processing
Impact dans le monde réel
15,000-word product page with pricing buried at bottom
Small AI model hits 8k token limit midway
AI never sees price, can't cite product
JSON-LD schema in <head> with all key facts
AI reads complete product data in 200 tokens
Perfect visibility, consistent AI citations