Attention Decay

Extended definition

Attention Decay describes how AI models' attention mechanisms prioritize information based on position, with attention strength typically decaying from beginning to end of documents. First paragraphs receive strongest attention; middle sections receive moderate attention; endings receive weak attention unless summary signals trigger reinforcement. Decay isn't linear—it varies by model architecture, document length, and content structure. Attention Decay means critical information buried in middle or end sections has lower extraction probability than identical information placed early. Decay also occurs across retrieved sources: first sources retrieved receive more attention than later sources.

Why this matters for AI search visibility

Attention Decay determines whether your most important information gets extracted or overlooked. Critical brand differentiators, key product benefits, or essential context buried deep in content may never reach the AI's generated answer even if factually present. For content optimization, Attention Decay demands 'most important first' structuring: lead with core claims, entity identification, and key facts. Burying crucial information in later sections—common in academic or narrative writing—virtually guarantees AI will miss it. Understanding decay patterns helps structure content for maximum extraction probability.

Practical examples

• A/B test shows entity mentioned in first paragraph gets extracted 87% of time versus 31% when first mentioned in paragraph 8
• Key differentiator positioned in concluding section gets cited only 12% of time versus 74% when repositioned to introduction
• Content restructured to 'front-load' critical facts increases citation rate 3.4x without changing total information