We measure what the AIs say. That is the whole company.

Who runs it

Saymetry is built and operated by Blackwood Management LLC. It is a small, independent, self-funded team — no investors, no board, and nobody whose interests compete with yours. You can reach a human at hello@saymetry.com, and the person who answers is the person who builds it.

The obvious conflict, named

We sell the measurement and we also sell the work it points at. So the fair question is whether we have an interest in your numbers looking bad.

The honest answer is that we would if the numbers were a judgement. They are not. A recommendation rate is a count of how many answers named you, out of answers you can read in the report, against questions you can see and edit. Every figure carries the interval that says how much to trust it, and when a sample is too small to support a rate we print that instead of a number. There is no dial to turn.

The scan is also priced to stand alone. Most people buy it, act on the plan themselves, and never speak to us again, which is a normal outcome rather than a failed upsell.

What we will not do

• Quote a number we cannot defend. Every figure carries its interval, and where a sample is too small to support a rate we say so instead of rounding it into something that looks confident.
• Invent a quote. Every sentence attributed to an engine is checked character by character against the raw answer it came from.
• Give you legal advice. We report what we can see on your site. Whether that meets a regulation in your jurisdiction is a question for a lawyer, and we will say so rather than guess.
• Promise you a result. Nobody controls what a model says, and anyone selling you a guaranteed place in an AI answer is selling something they cannot deliver. We can tell you what the engines say today, what changed since last time, and what the measured evidence says is worth trying.

How it works, in public

The formulas are published rather than described. The methodology page gives the recommendation rate, the share of recommendations, the confidence interval and the change rule, along with a plain list of what the measurement cannot tell you.