The words we use
AI visibility metrics, defined: what each number actually counts
Every vendor in this category throws around "AI visibility" numbers without saying what's in the numerator or the denominator. Here's what each one actually measures, and what it doesn't.
28 Labs · August 2026
AI visibility metrics are raw counts over a frozen set of questions: how often a brand gets named, how often it lands in the top three, how often it's picked first. Citations measure something different again - whether an engine read your page, not whether it recommended you. If a number can't be traced back to specific questions and specific answers, it isn't a measurement. It's a guess wearing a percentage sign.
The unit everything else is built on: the wave
Before any metric means anything, you need to know what it's a percentage of. That's the wave.
Wave
One complete measured pass of a frozen question set, asked across every AI engine you track, on a fixed date. The wave is the denominator behind every other metric on this page - Visibility Rate, Top 3 Rate, and #1 Pick Rate are all percentages of one wave's total answers. Change the questions, swap an engine, or move the date and you're comparing two different instruments, not two points on a trend line.
This is the discipline that separates measurement from marketing copy. A number without a wave behind it is a claim. A number with a wave behind it is re-derivable - anyone can re-ask the same questions and check your math.
The three naming rates
These answer three different questions, and mixing them up is the most common way "AI visibility" gets misread.
Visibility Rate
The share of answers, within one wave, where the brand is named anywhere in the response - any position, any tone. Numerator: answers that mention the brand at all. Denominator: total answers in the wave. It's the loosest of the three rates. A brand can post a high Visibility Rate and still lose every sale, because being named last, or named as a caution, still counts.
Top 3 Rate
The share of answers where the brand appears among the first three names the engine gives. Numerator: answers where the brand lands in position one, two, or three. Denominator: same wave total. Buyers reading a list rarely scroll past the first three - this is the rate that tracks whether you're in the part of the answer people actually read.
#1 Pick Rate
The share of answers where the brand is the engine's first recommendation - the name it leads with when someone asks "who should I use for X." Numerator: answers where the brand is named first. Denominator: same wave total. This is the revenue metric. The other two are leading indicators; this is the one that correlates with the buyer actually choosing you.
Being read isn't being named
Citation and Visibility Rate get confused constantly, and the confusion costs brands real diagnostic accuracy.
Citation
A citation means the engine read your page as a source while building an answer - it shows up in the response's linked sources or footnotes. It is not the same as being named. A page can be cited and the answer can still recommend a competitor by name in the visible text. Citation counts measure your performance as a source. The naming rates above measure your performance as an answer. Different instruments, different questions.
Grounding query
The internal search an engine writes for itself before it answers you - not what the user typed, but the reformulated query the model fires to retrieve live content. One user question can trigger several grounding queries behind the scenes, and your page gets cited when it wins one of them. We wrote a full breakdown of this in our review of
Bing Webmaster Tools' AI Performance report, currently the only place a major engine shows site owners its own grounding queries.
Two instruments you can't add together
AI-sourced vs. AI-influenced
AI-sourced is a visitor who clicked through from an AI answer and left a trackable trail - a referrer, a UTM, a link in the response. AI-influenced is a customer whose decision an AI answer shaped, but who arrived through a different channel entirely, because most AI answers today don't carry clickable links at all. One is measured by click data. The other is measured by survey or self-report. You cannot sum them - adding a click-tracked number to a survey-inferred number produces a total with no real denominator behind it.
Share of voice: relative, not absolute
Share of voice in AI answers
Your naming rate relative to competitors, measured across the identical question set in the identical wave. Numerator: your rate (Visibility, Top 3, or #1 Pick - specify which one). Denominator: that same rate summed across every brand named in the category, including you. Share of voice only means something when the comparison set is fixed. A "30% share of voice" tells you nothing until you know who else was in the room and exactly what was asked.
Watch for thisAny single "AI visibility score out of 100" that can't be traced back to specific questions and specific answers isn't a measurement - it's a blend of unknown weights on an unknown question set. Ask what questions were asked, how many, across which engines, on what date, and ask to see the underlying answers. If the vendor can't produce that trail, the composite score is opinion dressed as data.
What we do about it
Every metric we publish - Visibility Rate, Top 3 Rate, #1 Pick Rate, share of voice - re-derives to a frozen question set, a fixed engine list, and a dated wave. No blended score, no hidden weighting. If a client asks why a number moved, we point at the specific question and the specific answer that moved it.
That's the whole discipline: raw counts over denominators anyone can check, not a score we're asking you to trust on faith. For how this differs from what most AI visibility tools actually report, see what AI visibility tools can and can't measure.