Case study · publisher archetype walkthrough · 2026-05-19

Read Stacks: applying the audit to a 641-page book-summary publisher.

Read Stacks (readstacks.com) is a non-fiction book-summary publisher in the operator's fleet. It ships 45 books × ~24 chapter summaries each — 641 published chapter pages plus 45 book-level hubs and 6 curated stacks. It was built for human reading: long-form chapter summaries with editorial framing.

This page walks through what the Citation Readiness Score audit surfaces for a publisher archetype of this shape — the specific gaps to expect, the fixes that compound across the 641 chapters, and the honest unknowns (this is a walkthrough, not a fabricated before/after).

What the audit surfaces (per dimension)

When we run the Citation Readiness Score on a publisher archetype with Read Stacks's shape, the per-dimension findings follow a consistent pattern. Below is the audit's structural diagnosis — not a fabricated before/after number, but the gap categories the audit would identify.

• Dual Fit — the biggest gap. Chapter pages typically open with editorial paragraphs ("In this chapter, the author argues...") rather than a quote-extractable fact. LLMs that grab a 50-word paragraph for citation find prose-framing instead of substance. Fix: rewrite the first paragraph to lead with the chapter's core fact or claim, in ≤50 words.
• Entity Coherence — common gap on solo-operator publishers. Organization schema is usually present; Person schema for the operator/author is often missing. Adding Person schema with sameAs to LinkedIn + GitHub + Substack lifts the dimension materially. One JSON-LD block in the global layout applies site-wide.
• GEO Readiness — DefinedTerm schema for recurring concepts. Recurring concepts across summaries (book titles, author names, frameworks like "antifragile" or "flow state" or "quadrant II") benefit from DefinedTerm schema wrapping. LLMs extract these as terminology, then cite the source.
• SEO Foundation — usually already strong. Static export, mobile-perfect, schema present, sitemap + IndexNow wired. Title/meta/H1 structure typically clean. This dimension is rarely the bottleneck on a well-built publisher.
• Bot-Crawl Health — Cloudflare AI Crawl Control + llms.txt are the typical lift. Many sites that allow AI bots in robots.txt still have Cloudflare's AI Crawl Control set to "Block" (a default for zones added in 2025). Two-click flip in the dashboard + add /llms.txt at root and the dimension moves to 0.9+.

The fixes that scale to 641 pages

The publisher archetype's strength is that most fixes are template-applicable: edit the layout, the JSON-LD block, the global glossary — and 641 chapter pages inherit the fix simultaneously.

1. Global JSON-LD chain. One Organization + Person + sameAs block in the root layout. Applies to every page on the site. ~1 hour of operator work.
2. First-paragraph rewrite template. A consistent pattern: the chapter's core claim or fact, ≤50 words, no editorial throat-clearing. Apply to top-50 most-trafficked chapters first, then the long tail. This is the only per-page work — but it's the highest-leverage one.
3. Centralized DefinedTerm glossary. ~300 recurring concepts wrapped once in a shared glossary component, referenced wherever they appear in chapter text. Pages inherit DefinedTerm schema automatically.
4. Two CloudFlare clicks + one llms.txt file. The Bot-Crawl Health dimension lifts from 0.6-0.7 to 0.9+ with a five-minute change.
5. Set up CitationDesk polling on the top-50 chapters. Track citation events weekly across the four LLMs. The audit isn't the goal — sustained tracking is. The audit identifies the gap; the polling verifies the lift.

What we honestly don't know yet

Read Stacks is shipping. The audit-driven rewrites are queued on the operator's side. As of this case study's publication, we don't yet have a measured before/after Citation Readiness Score for the 641 chapters — and we won't fabricate one.

We also can't tell you whether ChatGPT or Perplexity actually cite those chapters. CitationDesk measures readiness, not observed citations — it does not poll AI engines. Answering that question needs a dedicated monitoring tool, and we would rather say so than imply a number we don't have.

Subscribe via RSS or watch the changelog — if we get a measured before/after for this site, it lands there.