You've seen the email. A tool scans your website, hands you an "AI visibility score," and — right on cue — a plan to move it up. On their platform. For a fee. Forever. The number feels official. It has a gauge, a color, a decimal. And it's designed to do exactly one thing: scare you into a subscription before you ask the only question that matters.
Can you reproduce it?
That's the whole test. A real measurement is one anyone can re-run and get the same answer — including someone who doesn't work for the company that sold it. A number you can't see inside, can't check, and can't reproduce isn't a measurement. It's a marketing line wearing a lab coat. A score you can't verify isn't a measurement. It's a leash.
How to tell a real score from a vanity one.
Four questions cut through every "free AI visibility score" in your inbox. A real one passes all four. Most don't pass one.
- Is it reproducible? Same website, same day → same number, every time. If two runs disagree, or if you can't run it yourself, it's a guess with a font.
- Is the method published? The score should be nothing more than a disclosed set of tests applied to your site. If you can't read the tests, the number is a black box — and a black box is where a vendor hides a thumb on the scale.
- Is it multi-engine? "AI visibility" isn't one thing. ChatGPT, Claude, Gemini, Perplexity, and Google's AI Overviews each answer differently. A single-engine number rounded into one grade hides where you're actually weak.
- Does it measure being named — or just mentioned? Getting your keyword to appear is not the same as an engine recommending you by name when a buyer asks who's best. Only the second one sends you customers.
The oldest trick in marketing, in a new coat: celebrate being mentioned while your phone stays quiet.
What PRAGMA does differently.
PRAGMA is the neutral certifier for AI visibility — and we resolve the trust problem the honest way. We don't ask you to believe our number. We publish the ruler and let you check the work. We're not the neutral party — we sell the fix, and we say so plainly. What we own is the neutrality of the method: the tests are disclosed, the result is reproducible, and you can re-run it yourself against a frozen baseline. Trust moves from "trust us" to "run it and see."
We ask ChatGPT, Claude, Gemini, Perplexity, and Google AI Overviews the exact questions your buyers ask, and record — per engine — whether your business is named.
We separate the pillars — answer-engine visibility, parametric (model-memory) visibility, and local presence — because a single blended grade hides which one is failing.
We freeze a baseline before we touch anything, so any change is provable, not asserted. And we don't claim a score predicts revenue — we report what the engines actually say. That's the difference between a ruler and a leash.
The proof isn't a promise. It's a receipt.
Anyone can claim their score is honest. Here's ours, run on a real business — our own founder's, which is exactly why we could open every number.
Brand 9 Signs, a 40-year signage company, was effectively invisible to AI in its own market. We ran the disclosed check, froze the baseline, answer-shaped its pages on its own domain, and re-ran the exact same check. Gemini, Perplexity, and Google's AI Overviews now name Brand 9 Signs when buyers ask who's best in its category. Same test before and after — the movement is on the record, measured against the market, not a vanity impression. We even show the pillars that are still weak, because hiding them would make our own score a vanity metric.
That's the standard. If a "free score" in your inbox can't show you its tests, can't be re-run, and can't tell you whether AI actually names you — you already have your answer about what it's worth.
Are you the answer?
See whether AI names your business today — with a check you can run yourself, not a score you're asked to trust.
Are you the answer? Find out