The playbook

What this document is, how it was produced, what "verified" means here, and the fastest path through it depending on how much time you have.

4 min read · reviewed 2026-07-26

The base layer everything else assumes. One origin, one host, one trailing-slash policy, real status codes. Get these wrong and nothing built on top of them can work.

6 min read · reviewed 2026-07-26

Titles, descriptions, canonicals, robots directives and social cards. The cheapest wins available and the most commonly half-finished layer on an otherwise well-built site.

7 min read · reviewed 2026-07-26

JSON-LD that describes a page to a machine. What the 42 types are for, which ones earned their place, and why optional fields must be omitted rather than emitted empty.

5 min read · reviewed 2026-07-26

One script, one graph, stable identifiers. How to describe 2,100 pages as one entity rather than 2,100 unrelated documents, and the test that proves the references resolve.

4 min read · reviewed 2026-07-26

robots.txt generated from code, private routes that stay private, orphan detection, and IndexNow. Controls what gets fetched, how often, and by whom.

4 min read · reviewed 2026-07-26

The robots.txt decision nobody makes deliberately. Which crawlers return traffic or citations for the bandwidth they consume, crawler by crawler, and why the AI-versus-search framing is the wrong one.

5 min read · reviewed 2026-07-26

Differentiated lastModified, a priority ladder that says something, hreflang per URL, and an RSS feed generated from the same source so the two cannot disagree.

4 min read · reviewed 2026-07-26

How to generate a thousand pages without generating a thousand doorway pages. The strip-the-entity-name test, per-entity numbers, computed FAQs, honest fallbacks and sibling linking.

6 min read · reviewed 2026-07-26

The single control that stops a dynamic route from generating an unbounded number of thin pages. Gate against an authoritative list, 404 everything else, and treat the list as data.

4 min read · reviewed 2026-07-26

Author identity that resolves, freshness dates that are actually maintained, claims traceable to evidence, and why publishing your own failures is the strongest trust signal available to a new site.

5 min read · reviewed 2026-07-26

The newest optimisation surface and the least codified. How answer engines select and quote, why self-contained sentences beat pronouns, and how to measure whether any of it worked.

6 min read · reviewed 2026-07-26

The two files, what belongs in each, and the blocks that do the real work: key facts, negative statements, the content index and citation guidance.

4 min read · reviewed 2026-07-26

LCP under 1.8 seconds on mobile, and the five things that actually cause a content site to miss it. Inlined CSS, self-hosted fonts, modern image formats, code splitting and static-first rendering.

5 min read · reviewed 2026-07-26

Security headers including a real CSP and the compromise it requires, the API validation pipeline, timing-safe secret comparison, and the accessibility work that is also SEO work.

6 min read · reviewed 2026-07-26

The mobile-first verification pass, why 375px is the width that matters, and the four defects that make a technically correct site feel broken on a phone.

3 min read · reviewed 2026-07-26

How to tell whether any of this worked. Query segmentation by page tier, distinguishing an indexation problem from a ranking problem, and what to look at weekly versus quarterly.

5 min read · reviewed 2026-07-26

What to build first with two weeks instead of six months, ordered by expected return, using the effort and impact values from the check registry rather than intuition.

4 min read · reviewed 2026-07-26

How each mechanism in this playbook backfires. Over-indexation, FAQ schema treated as spam, canonical loops, key leakage, doorway classification, and a CSP that silently breaks the funnel.

5 min read · reviewed 2026-07-26

The honest list. Every gap in the reference implementation and in this site, what each one costs, and how to fix it. Not sanitised, because a document that claims completeness is not checkable.

5 min read · reviewed 2026-07-26