Updates

This book documents an experiment that is still running. Every future edition is included in what you already paid: no upgrade, no second purchase. Current edition 1.3, data through 2026-08-05.

What is coming

Edition 1.2: The plateau verdict

The book ends with the site stuck at position 8–10 and the argument that only editorial links move it. The next edition reports whether twenty to thirty new referring domains actually broke the plateau, or did not.

Edition 1.3: The domain decision, resolved

Chapter 18 declines to assert that the country-code domain causes the geographic ranking gap, because the data does not cleanly support it. A migration would settle the question. Whatever happens gets written up, including a null result.

Ongoing: Dataset refresh

Every edition ships the current full export, so the numbers in the book can always be checked against the numbers on the site.

Change log

v1.3: A chapter on naming, and four chapters made deeper

Chapter 9 explained what a name does, 44.1% of clicks. It never explained how to choose one, which is the decision that made this project lose its own brand query to a near-identical competitor. Chapter 10 is the method, written up with the candidates that failed and why.

• Chapter 10, How to Choose a Name You Can Own: the three tests, seven real candidates killed by evidence, RDAP verification, and the TLD renewal trap
• Chapter 4 gains a discovery method, how to actually find the queries nobody serves, and how to judge the weakest result on page one
• Chapter 14 gains the mechanism behind the settlement delay, a way to measure your own lag, and the exception: a cliff is technical and does not wait a quarter
• Chapter 15 gains the ±2σ calculation for your own noise threshold, and why smaller sites are noisier as arithmetic rather than misfortune
• Chapter 20 gains the segmentation procedure and a sample-size floor, the tablet row is the best on the site and 1% of impressions
• Now 30 chapters and 6 appendices, 49,432 words

v1.1: Two new chapters: answer engines in practice, and a month of zero

Chapter 12 covers what was actually done to build visibility inside AI assistants, and why the referrer log cannot measure it. Chapter 21 documents 31 consecutive days in which the second search engine sent exactly zero traffic, and the diagnosis that stopped a wrong lesson being learned from it.

• Chapter 12, How to Be in the Training Data: the platforms, the community rules that get founders banned, and a monthly prompt-set protocol that is the only honest scoreboard
• Chapter 21, Thirty-One Days of Zero: 128 days of second-engine data, a binary deindex with no notification, and a fix that took minutes once it was finally tried
• New finding: the second engine was 20.59% of Google’s clicks in April, and was registered months late
• Chapter 15’s falling-CTR thesis independently reproduced on a second index, 46.56% CTR on 59 impressions/day became 10.68% on 149.6
• Ten chapter openings rewritten, and a channel-concentration exercise added

v1.0: First edition

Thirty-six chapters and six appendices on taking one product from zero to 100K+ users on search alone, with nothing spent on ads.

• The complete dataset behind every claim, reproducible from the exports in Appendix A
• Three findings that contradicted the author’s own assumptions
• Fourteen downloadable templates, including an AI-agent brief and a working page generator
• An in-browser audit tool that runs every diagnostic on your own export
• Five dated, falsifiable predictions (Appendix F)