AI NativeThe two numbers at the heart of AI Native are mention rate and recommendation rate, and they are not the same thing. Confusing them is the most common mistake teams make when they first look at AI visibility, so this guide pulls them apart.
Being named is not being chosen
A mention means the answer named your brand. The assistant said your name. That is it.
A recommendation means the answer put you forward as a choice. The assistant suggested you, framed you as a good option, told the buyer to consider you.
Those are different events. An answer can mention you while recommending a competitor. Imagine a buyer asks which provider to choose, and the assistant replies: "Brand A is the strongest option for most people. Brand B and Brand C also exist." Brand A was recommended. Brands B and C were merely mentioned. All three have a mention; only one has a recommendation.
This is why AI Native scores the two separately instead of folding them into a single visibility score. A single score would let a strong mention rate hide a weak recommendation rate, and the weak recommendation rate is usually the real problem.
How AI Native tells them apart
For every answer a scan reads, AI Native decides two things: did this answer mention the brand, and did it recommend the brand. Because each prompt is run more than once and AI answers vary between runs, these become rates rather than yes-or-no facts. Your mention rate is the share of answers that named you; your recommendation rate is the share that put you forward. Running and averaging is what makes the rates stable enough to compare over time.
The gap is the work
The interesting figure is the distance between the two. When your mention rate is high but your recommendation rate is low, AI knows you exist but is not choosing you, and that gap is your work list. It points at a specific failure: you are visible but not preferred, present in the answer but losing the decision inside it.
Closing that gap is different work from getting mentioned in the first place. Getting mentioned is about being known to the model at all. Getting recommended is about the evidence the answer leans on, the sources that make the assistant comfortable putting you forward rather than a rival. The scan exposes those sources, which is where you start.
Reading them together
- Low mention, low recommendation: you are largely invisible. Start by getting into the conversation.
- High mention, low recommendation: known but not chosen. The classic gap. Focus on becoming the recommended option, not just a named one.
- High mention, high recommendation: you are winning the answer. Hold it, and watch the trend so you notice if a competitor erodes it.
Both numbers belong on the same screen for exactly this reason. Read alone, either one misleads. Read together, they tell you whether your problem is being seen, being chosen, or neither.