Introducing The AI Citation Institute: we measured ourselves at zero, and published it

The AI Citation Institute is an instrument. It runs the real questions buyers ask an AI assistant — ChatGPT, Claude, Gemini, Perplexity, Google AI Overviews — against a frozen panel, samples each query several times, and reads which brands the answers actually name. That reading is the product: not a hunch about whether AI mentions you, but a measured share of answers you can watch move.

We ran the instrument on ourselves first

On our day-0 scan, in the vendor-selection answers for our own category, The AI Citation Institute appeared in 0.0% of them. The established tools took the rest — Semrush at 59.4%, Profound at 48.4%, Otterly.AI at 43.8%. That zero isn't a confession we buried; it's the honest starting point the whole concept predicts for a brand nothing citable on the open web points to yet.

The instrument is only honest if it grades us too. So we publish our own reading on the same panel we point at clients — live, and never retouched.

What we publish

Two research records go out with the launch. One is a 618-trial experiment on voice: a plain, casual-precise register beat a corporate one for citation, by 33 percentage points on Claude, and numbers written sloppily were reproduced 0% of the time. The other is an audit of all 91 of our own posts across four engines: pages with zero inbound internal links were cited 4.5% of the time, and pages with ten or more, about 44%. Both are open access, with the methodology and the honest limitations attached.

The scoreboard holds the live market reading and our own climb from zero. The research holds the studies behind the levers. That's the bet: measure it, act on it, and show the receipts — including our own.