LLM引用文献
LLM Citation is the technical mechanism by which a Large Language Model attributes specific generated text to a training document or retrieved data source. For brands to earn citations, content must be formatted (often via JSON-LD) so the model's attention mechanism recognizes it as the primary authority on a topic.
The Technical Foundation of AI Visibility
There's a critical difference between "training data" (background knowledge the LLM absorbed) and "cited sources" (active references it displays to users). When ChatGPT says "Nike makes shoes," that's general training data—no citation, no traffic. When it says "Nike released the Air Max DN on March 26, 2024" and links to the press release, that's a citation—you get the click. The technical key is structured data: JSON-LD schema tells the LLM's retrieval system exactly what information to extract and attribute. Without structured markup, your content becomes generic training fodder. With proper implementation, you become a cited authority that drives measurable traffic from AI interfaces.
General Mention vs. LLM Citation
現実世界の影響
Product page has unstructured pricing information
Perplexity.ai: "Prices vary, check their website"
User clicks competitor with clear pricing
Add JSON-LD Product schema with price + availability
Perplexity.ai: "Product X is $99, in stock" [citation]
User clicks citation, high purchase intent