As LLMs from OpenAI, Anthropic, and Google become the main interface for information, the role of the website is shifting fast. Users get answers directly from models without visiting the source, but the bigger issue is hallucination. LLMs often summarize outdated pages, merge multiple websites, or generate confident but incorrect details because websites aren’t structured for retrieval or authority signaling.
If hallucination is partly a data-structure problem, should we start optimizing websites specifically to reduce it, with structured AI-readable summaries, versioned knowledge endpoints, and embedding-ready content? Maybe the next evolution isn’t SEO, but hallucination optimization. Curious what real problems others are seeing and how you’re adapting.