Measured, or merely asserted?
The AI-visibility category has learned the vocabulary of measurement faster than the discipline. Six patterns to check before you trust a number, whoever shows it to you. Including us.
Asserted
A dollar-value metric with no visible derivation. If nobody can show the arithmetic from measurement to dollars, the number is a decision aid for buying, not for acting.
Measured
Every number links to what produced it: the prompts run, the answers returned, the sample count, and the interval around the result.
Asserted
Provenance vocabulary without math: tiers named measured, estimated, computed, with no sample sizes or intervals anywhere behind the labels.
Measured
Provenance labels backed by the statistics they imply: a measured share ships with its n and its 95% Wilson interval, every time, or it ships as NOT MEASURED.
Asserted
A confident score in the first minute. Real sampling takes time; an instant verdict is a model wearing a measurement's clothes.
Measured
Numbers that start with wide bands and visibly narrow as samples accumulate, with change flagged only once it clears a significance test.
Asserted
Capability built by a partner, marketed as first-party once the press cycle fades. What is quietly true about the plumbing tends to be quietly true about the numbers.
Measured
A public methodology page that says exactly what is measured, how often, on which engines, and what the tool cannot see.
Asserted
Screenshots of dashboards with no account attribution, and month-12 projections styled as case-study outcomes.
Measured
Only measured history, clearly dated. A projection is a model; a case study is a record. Confusing them is the tell.
Asserted
The same vendor writes your content, places your links, and grades the results. The grader has a stake in the grade.
Measured
Measurement separated from fulfillment. We publish nothing and place nothing, so a bad number costs us nothing to show you.
A pattern to check for before you pay for any AI visibility score
A genre of one-shot report is spreading through this category, and it has a recognizable shape: a struck-through price with no history behind the higher number, a 1 to 100 score with no published methodology, no sample size and no interval, results that appear only after checkout, and testimonials with precise outcome figures on a product too new to have customers. Any one of those can be innocent. All of them together is a score being sold, not a measurement.
The checks work on anyone, including us: is the methodology public, is the sample size printed on the number, can you see a real result before paying or identifying yourself, and does the vendor publish what its own instrument gets wrong? Ours are at how we measure, the measured noise floor and the free, ungated checkup.
Hold us to the same bar: our methodology, with the math shown, and a free audit that runs a real measurement before asking you for anything.