Chapter 2 — Key Takeaways
The Ranking Algorithm: What We Know, Think We Know, and Are Guessing. A one-page card. Skim before an audit; reread when someone hands you a confident "ranking factor."
The core claims
- Ranking is a contest, not a checklist. You don't complete factors; you out-answer competitors for a specific query, judged by systems that weigh many signals differently each time.
- "200 ranking factors" is folklore. It hardened an old, round, illustrative Google figure ("more than 200 signals") into a fake, fixed list. Any "complete list" mixes confirmed items with guesses.
- Confirmed ≠ powerful. Several confirmed signals (HTTPS, Core Web Vitals) are real but lightweight. Confirmation says a signal exists, not that improving it moves rankings.
- Relevance is the biggest confirmed lever, by far. Everything else is in support of matching intent.
- No one can read the rulebook — not even Google. Modern ranking is machine-learned systems whose behavior is emergent; there is no readable sentence explaining why A beats B.
- The deliverable is a habit: sort every claim by evidence tier before acting on it.
The ranking-claim map
| CONFIRMED (build on it) | LIKELY (act, hold loosely) | DEBUNKED (stop) |
|---|---|---|
| Relevance / intent match (biggest) | Engagement / click signals | Meta keywords tag |
| Links (as a signal) | Topical authority | Keyword density % |
| Mobile-first indexing | Freshness as a blanket habit | Word-count targets |
| HTTPS (lightweight) | "Comprehensiveness helps" | Domain age |
| Core Web Vitals (tiebreaker) | "Post daily to rank" | |
| Language / locale | Third-party "authority" as a signal Google reads | |
| Freshness (query-dependent) |
The evidence ladder (never let a claim exceed its rung)
| Rung | Kind of claim | Example | How to treat it |
|---|---|---|---|
| 1 | Confirmed by Google | "HTTPS is a lightweight signal." | Build on it |
| 2 | Strong correlation / evidence | "More referring domains ~ higher rank." | Act, hold loosely |
| 3 | Practitioner experience | "This fix worked on my site." | One data point |
| 4 | Speculation / sales pitch | "The secret 2% density rule." | Ignore as a basis for spend |
Three questions for any SEO claim
- Which rung is it on? (Confirmed / correlation / anecdote / guess.)
- What would prove it wrong? (Unfalsifiable ≠ knowledge.)
- Who benefits if I believe it? (Follow the incentive.)
Reading a correlation study without being fooled
- Correlation or causation? (Almost always the former; the headline implies the latter.)
- Could a confounder explain it? (Quality/authority driving both.)
- Could the arrow point backward? (Ranking causes links, engagement, brand searches.)
- What's the sample, and who's selling the conclusion?
The machine-learning trio
| System | Year | What it changed |
|---|---|---|
| RankBrain | 2015 | First ML ranking system; interprets query meaning, esp. novel queries |
| BERT | 2019 | Reads words in context — meaning over string-matching |
| MUM | 2021 | More powerful/multimodal; real, but a vague/narrow live-ranking role |
The uncertainty to carry: the AI Overview
A page can rank #1 — doing everything the confirmed signals reward — and still lose clicks to an AI Overview answering the question on the results page. Ranking well is necessary, no longer sufficient. Build depth, originality, and traffic diversification. (Home: Chapter 36; content response: Chapter 13.)
On your own site tomorrow
View Page Source on one page: delete any <meta name="keywords"> fossil, confirm viewport and https
are present — then stop optimizing debunked factors and put the freed time into relevance.