Case Study Zero
The experiment runs on me first.
The fair test of a measurement is whether you will point it at your own business and publish whatever it finds. So my own site is a standing subject: a locked seven-query scan, run on patrickrobinson.consulting, in public, from a measured start. This public tracker is its own seven-query measurement contract; the paid Audit uses a separately defined, larger fixed scope.
Real before-and-after only. No sample data, no numbers a chatbot made up for a slide. The baseline below is a real zero, and it can't be quietly edited later.
Citation rate today
0/7
category buyer queries where an AI engine currently cites my work, across Perplexity, Claude, ChatGPT and Gemini.
Baseline measured . This number is the experiment. Every rescan is published here, whatever it shows; the next scheduled verdict is 17 August 2026.
The series
Episode 01
The honest baseline
The opening baseline, recorded 6 June 2026: cited in zero of seven category buyer queries, published before the result was known. The transcript is preserved as recorded, with a dated correction below it.
Read the transcript
Correction, dated 17 July 2026. This transcript is preserved exactly as recorded on 6 June 2026. Three things it says no longer reflect how I describe this experiment. First, the thirty, sixty, ninety day sequence and the "fast wins" were an early hypothesis, not a result; the engines control citation outcomes and the timetable, and nothing here should be read as a promised sequence. Second, "the same audit I run for clients" overstated it: this public tracker is a locked seven-query measurement contract, and the paid Audit uses a separately defined, larger fixed scope. Third, running the method on my own site is an open experiment, not proof the method works. The baseline, the queries and the measurement rules are unchanged, and the number above reports every rescan whatever it shows.
I refuse to sell a method I haven't run on my own business first. So before I quote anyone a price, my own site is the first subject.
This is Case Study Zero. The Citation Engine, running on patrickrobinson.consulting, in public, from a measured start.
Here's why it matters. When a buyer asks an AI for a recommendation, being visible means being one of the sources the answer cites. That is the whole game now.
The baseline is real. I ran the same audit I run for clients. Seven buyer-intent queries, the questions a founder would actually ask, across Perplexity, ChatGPT, Claude and Gemini. The result? Cited zero times, out of seven. I published that number on the sixth of June, before I knew where it would land, where it can't be quietly edited later. That's the point.
Now the fix, in the open. The same thirty, sixty, ninety day sequence I hand to clients. First, surface presence. The places AI engines already pull from. Fast wins that get you on the board. Then citation-shaped depth. Long-form, founder-bylined essays with specific claims and the reasoning shown. Not blog posts chasing clicks. Pages built to be quoted.
Every movement in that number gets an episode. Not projections. Not sample data. A real before and after, or nothing. You don't have to take the method on faith. You can watch it run. And if you want it pointed at your company, it starts with an audit. What will your baseline reveal?
Episode 02
Upcoming
The first verdict
Publishes after the 17 August 2026 rescan, whatever the number shows. A movement gets reported as a movement; a zero gets reported as a zero, with the post-mortem in the open.
Locked until the 17 August rescan. No teaser, no placeholder result. It publishes whatever the verdict is.
1 of 2 episodes published. Episode 2 publishes after the 17 August rescan, whatever it shows. Later episodes use declared evidence events, not a promotional calendar.
The rules of this experiment
A separate, locked public contract. This baseline uses seven buyer-intent queries across four answer engines, run multiple times. It is not the paid Audit, which uses a separately defined, larger fixed scope. The before is a real zero.
The work is a hypothesis, not a formula. The original plan described a thirty, sixty, ninety day sequence with fast wins. I no longer describe it that way: the engines control the outcome and the timetable. The work itself is surface presence plus deeply sourced, founder-bylined pages; whether and when that translates into citations is exactly what this experiment measures. The essays on this site are part of that depth.
In the open, where it can't be edited later. Each declared rescan documents what changed, including no change. The number was published before the result was known and is not revised around the outcome.
The same measurement harness also powers my original research: Who AI Search Cites When Buyers Ask for AEO Help analyses 649 classified citation appearances across these four engines, with the full dataset and frozen methodology downloadable.
Want your own baseline measured?
The paid Audit applies the same measurement discipline at a bigger, fixed scope: a declared query set across the four engines, named competitors compared under the same rule, and the evidence and priorities shipped as a deck. What you do with the findings is agreed one month at a time.