We measured ourselves. We were named in none of seventy-eight answers.

The questions a buyer would type. How to find out what AI assistants say about a brand. Who measures AI visibility. Tools for tracking brand mentions in ChatGPT. The ordinary discovery questions in our own category.

Every assistant named other companies. None named us.

Partly because it would be absurd not to. A company that sells measurement and will not measure itself is asking for a trust it has not earned.

But mostly because it is the only honest test of the thing we tell clients. We say the answer layer is earned through evidence in third-party sources, that it takes months rather than weeks, and that nobody controls it. If that is true, it should be true of us — and the only way to demonstrate it is to publish the starting line before doing the work, and then publish what happened.

Three things, none of them flattering.

We were competing with our own name. A domain we do not control was cited more often than our own — 46 citations against our 37. A search for our exact domain returns an unrelated biosignal research platform, two similarly-named sites and a large clothing retailer before it returns us.

We had published almost nothing. Dozens of frozen measurements sat in internal files with no public surface. We had produced the evidence and then not made it readable by the systems we measure.

And nothing was wrong with the website. Our robots file allows every AI crawler. No blocking, no noindex, no bot gating. The site was open and empty, which is the harder problem, and exactly the one we diagnose in others.

The baseline is frozen at 15 July 2026. We are re-running it against the identical question set, and whatever it says will be published here — including if the answer is that nothing moved.