Semantic AI Search.

Classic websites were often built as destinations: a person clicks, reads a page, and forms an impression directly. That still matters. But more discovery is becoming AI-mediated. A search summary, answer engine, assistant, or agent may encounter the site first, extract a passage, compare entities, or use the page as evidence before a human ever opens it.

That changes the design problem. A page must still be human-readable, but it also has to make its meaning explicit enough to be retrieved, summarized, cited, and distinguished from related entities.

The approach here is to treat a personal or expert website as a small public knowledge graph. Each important thing gets a stable source-of-record page: the person, the system, the invention, the lab, the concept, the note. Relationships are written in visible HTML, supported by internal links, and mirrored conservatively in structured data.

The goal is not to replace writing with schema. The goal is to make the writing legible: to humans first, and to machines without forcing them to guess.

yoavfekete.com is the live example of this method. The site is built around Yoav Fekete as the root person entity, then separates the connected entities instead of collapsing them: Clause, NaadLabs, Harmonic Sitar, YoYo Sitar, Afro-Raga, Representation, and the technical domains each have distinct roles and canonical pages.

The entity map makes the relationships visible. The concept page defines representation as a way of seeing. The first note, The Right Map for the Question, explains why representation changes what becomes visible, solvable, verifiable, or playable. The system pages show how the same principle appears in Clause, Harmonic Sitar, and NaadLabs.

This is not presented as a consulting page. It is a working example: a public person site structured so humans, search engines, and AI systems can understand the entities, follow the relationships, and cite the right page for the right claim.