Chapter 4 — Key Takeaways

A one-page card. Skim before the quiz; keep it beside you when you plan content.

The core claims

  • Things, not strings. Google moved from matching character strings to understanding entities and the meaning behind a query. A page can rank for words it never uses; a keyword-stuffed page can rank for nothing.
  • Cover the topic; don't repeat the keyword. Relevance is now about meaning and coverage, not repetition. This is the modern content game in one line.
  • The math already said so. BM25's saturation (repetition has diminishing returns) and length normalization (longer isn't automatically more relevant) are the mathematical reasons keyword stuffing and "longer is always better" stopped working.
  • Topical authority is real but unconfirmed as a score. Covering a subject broadly and deeply earns subject-specific recognition — a strong working model backed by partial confirmation, correlation, and experience, not a published dial.
  • You earn entity understanding by corroboration, not declaration. Be consistent, be described accurately in many trustworthy places, and Google's model of you strengthens. You cannot force a knowledge panel.

The mental models

Old model (strings) New model (things)
Does the page contain the query words? Is the page about the same entity and need?
Repeat the keyword; hit a density Cover the concepts an expert would; write naturally
One page per keyword phrase One deep page earns a cluster of related queries
Longer + more mentions = more relevant Focus and genuine coverage win; repetition saturates

How Google connects content to entities (three channels)

  1. Context — surrounding words disambiguate and map your page to the right entity.
  2. Co-occurrence — related concepts that appear together across the web signal genuine depth (learned from the corpus, not a word list to paste on).
  3. Structured data (JSON-LD) — states the entity outright; least ambiguous (full build: Chapter 18).

The two formulas, for intuition only

  • TF-IDF — a term matters when it's frequent here but rare across the web ("the" ≈ zero weight; "anode rod" carries real weight).
  • BM25 — TF-IDF plus saturation and length normalization; the default in major search libraries. Real IR math — not Google's confirmed algorithm.

Myths this chapter retires

  • 🚫 "Add LSI keywords to rank." — Not a real thing; LSI isn't how Google works. Cover the topic instead.
  • 🚫 "Hit the ideal TF-IDF density." — Saturation means extra instances barely help; there is no magic density.
  • 🚫 "Add sameAs and you'll get a knowledge panel." — Schema clarifies; it doesn't manufacture standing.
  • 🚫 "Topical authority is Domain Authority for a topic." — DA is a third-party link score, not a Google metric; topical authority is subject-specific recognition.

Key terms

entity · Knowledge Graph · knowledge panel · semantic search · topical authority · TF-IDF · BM25 · co-occurrence · disambiguation · sameAs

Do this on your own site tomorrow

  1. Search your brand name; read your knowledge panel (or note its absence) and your "People also search for."
  2. Pick one important page; list the concepts an expert would cover, and fill the gaps — without adding one more instance of the target phrase.
  3. Rewrite your About page's first two sentences so the entity — who, what, where — is unmistakable.
  4. Map your business as entity clusters (topics → equipment/problems/tasks), not a flat keyword list.