When people talk about "the algorithm," they usually mean a single black box. In reality, the infrastructure that feeds AI systems their understanding of entities operates as three distinct layers, each doing a different job. Understanding these three layers — and how they connect — is the foundation of entity engineering.

Layer one: the knowledge base

This is the structured data layer. It's machine-readable, organised around entities rather than pages, and it's where things exist as defined records with properties and relationships — this organisation is a Q-number, that person was born on this date, this product is manufactured by that company. Every major AI system, search engine, and knowledge panel draws on structured sources like this directly. It is not designed for humans to read casually; it's designed for machines to query.

Layer two: the evidence layer

The knowledge base alone isn't enough — systems also need corroborating, human-readable context to understand and describe an entity accurately. This evidence layer consists of descriptive content and media that explain what something is, in a form both humans and AI systems can interpret. It gives the structured records credibility and texture — turning a bare set of properties into something that can be summarised, explained, and cited.

Layer three: the alignment layer

This is the connective tissue — structured data on your own website that explicitly states "this page, and this entity in the knowledge base, are the same thing." Without this layer, the other two layers can exist in complete isolation from your site. The knowledge base might know about the entity; your website might describe it beautifully — but nothing tells any system that they're connected.

Why all three matter: a knowledge base entry with no evidence layer lacks context. An evidence layer with no knowledge base entry has nothing to corroborate. And either of those without alignment never connects back to your website at all. AI visibility depends on the chain being complete — not on any single layer in isolation.

Why almost nobody has built this

Each layer sits in a different "department" of how the web normally gets built. Web developers handle schema (alignment). Content teams handle articles and media (evidence). And the knowledge base layer isn't something most workflows touch at all — it's not a CMS feature, not a plugin, not a checkbox in an SEO tool. It requires a deliberate, separate process. That's the gap entity engineering fills.