About
The decision moved. Measurement has to move with it.
For twenty years, being found meant a position on a results page. Now a buyer describes their situation to an assistant and reads a paragraph. snowdust.ai exists to measure that paragraph — who it names, who it recommends, and what would have to change for that to be you.
Why we built it
Nobody could answer a simple question.
The question was: when someone asks an assistant what to use, does it say us? Not "do we rank" — that was answerable. Not "is AI mentioning us" — that was answerable and turned out not to mean much. The one that mattered had no instrument behind it, and the tools that claimed to answer it were counting something else and calling it the same thing.
So the product is narrow on purpose. It measures how the major answer engines — ChatGPT, Perplexity, Claude, Google AI Overviews, Google AI Mode, Google Gemini and Microsoft Copilot — describe, compare and recommend a brand across a fixed set of prompts; it explains why a competitor gets picked instead; and it turns that into fixes specific enough to hand to whoever owns the page. That is the whole of it. There is a great deal we do not do, and we would rather be clear about the edges than vague about the middle.
How we work
Four things we will not trade away
These are the commitments that shape what we build and what we refuse to ship. They are also the ones most likely to cost us a deal, which is roughly how you know they are real.
-
Every number opens
A score you cannot click into is a score you cannot argue with. Anything we report traces back to the answers it came from, so you can read the sentence your brand appeared in and disagree with us if we got it wrong.
-
Mentioned is not recommended
Being named in an answer and being the thing it tells someone to do are different outcomes, and blending them into one percentage hides the only movement that matters. We keep them apart.
-
The prompt set is a decision
Visibility is measured over the questions someone chose to track. That makes it easy to flatter, so we write the set down, keep branded questions separate, and treat editing it as a change to the instrument rather than a result.
-
We would rather say we do not know
This field is full of confident numbers that trace back to nothing. When the evidence is thin or the studies disagree, we say so — including when that is less convenient for us.
See for yourself
Read what we publish before you believe what we sell.
Our research posts are the honest test of whether the principles above are real. Every figure in them names its study, its sample and its date. Several widely repeated statistics are missing on purpose, because we could not trace them to a primary source — and we say so on the page rather than quietly using them anyway.
One of those posts argues that a visibility score is easy for a vendor to flatter, which is not an obviously commercial thing for a vendor to write. We would rather you know the failure modes of the category you are buying into.
The quickest way to judge us is still the free 14-day trial — a real reading on your own brand, and enough to tell whether any of this is worth your time.