llmo

"LLMはどのサイトにも書いてある汎用情報より、そのサイトにしかない一次情報を優先的に引用する傾向がある。"

"When using multiple LLMs, the most straightforward approach is for each app to directly hold the required provider's API keys and call them using their respective SDKs. However, as the number of providers increases, this 'hold directly, call directly' structure creates friction in various places."

"From the same FAQ content: ├── For Google: JSON-LD FAQPage structured data ├── For Users + Google: Displayed page FAQ (HTML) └── For AI Search: llms.txt / llms-full.txt"

"AI uses the Retrieval-Augmented Generation (RAG) method, referencing structured data and Google Business Profile information to generate answers. In other words, whether markup compliant with Schema.org is accurately implemented is directly linked to exposure in AI search."

"I’m looking to put together a serious AI study group"

"AI is 'known' based on whether it was included in the learning data, and how many contexts it was mentioned in, and not by SEO rankings."

"This book systematically explains LLMO (Large Language Model Optimization) that engineers can put into practice starting today, focusing on these 'three pathways'."

"LLMO is simply a 3-step process: Readable: Crawlers and AI can retrieve the page (technology), Extractable: Information is written in a form that is easy to use for answers (structure/wording), and Believable: Who wrote it, when, and on what basis is clear (trust/operation)."

"By 2026, over 80% of companies are predicted to deploy generative AI applications."