Average Position

In traditional search, position meant your rank in a list of blue links, and users scanned from the top. In AI search there is no list. The model writes a single, flowing answer, and order still carries meaning, just differently.

When a model answers "what are the best options for X," the brand it names first is doing more work than the brand it names sixth. The first-named option often reads as the default recommendation. Later names read as alternatives, caveats, or "also worth considering." A reader skimming a generated answer absorbs the opening far more than the tail, and increasingly an AI agent acting on that answer may only carry forward the first one or two names.

So position is a proxy for how central the model considers you to the question. Being named is table stakes. Being named first is a materially stronger outcome.

Average Position captures one thing precisely: your typical rank order within the answers that mention you. That precision comes with boundaries worth stating.

• It only counts answers where you appear. A brand mentioned in just 10% of prompts can still post a strong Average Position in those few answers. The metric says nothing about how often you show up; that is Share of Answers.
• It measures order, not tone. Being named first tells you the model leads with you, not that it speaks well of you. Sentiment is a separate axis.
• It measures order, not share. Two brands can both average "second" while one is mentioned far more often than the other.

Because of that first boundary, Average Position is dangerous to read alone. A brand can chase a beautiful Average Position by appearing in a handful of narrow prompts while remaining invisible everywhere else. Always pair it with a presence metric.

The honest way to use Average Position is as the depth companion to Share of Answers's breadth. Together they sort brands into useful shapes:

• Wide and high. Appears in most answers and usually named first. Category leadership in AI answers.
• Wide but low. Appears everywhere but always near the end. Known, but treated as an afterthought. The opportunity is to move up, not just to show up.
• Narrow but high. Named first in a small set of prompts. Strong in a niche; the task is to broaden.
• Narrow and low. Rarely appears and buried when it does. The hardest starting point.

Those quadrants turn two numbers into a strategy. Improving your position is a different job from improving your reach, and knowing which one you lack prevents wasted effort.

Average Position is one of several inputs into a full AI Visibility picture, sitting alongside presence, share, and sentiment. On its own it is a sharp, specific signal about prominence within the answer. Rolled together with the others, it helps explain not just whether models talk about you, but whether they lead with you. Prominence AI tracks Average Position across every major engine so you can see where you sit inside the answer, not only whether you made it in.

Is a lower Average Position number better?

Yes. Position is measured by order, so first place is 1 and lower numbers mean you are named earlier. An Average Position of 1.8 means you are typically among the first brands mentioned; 7.2 means you usually sit near the end of the list.

Does Average Position work the same across every AI engine?

The concept is the same, but results vary by engine because each model structures answers differently. That is why it is worth tracking position per engine rather than assuming one number represents them all.