What is Average Position?
Average position is the mean rank at which your brand appears in AI answers that name several brands, where position 1 means you are mentioned first. It shows not just whether AI assistants recommend you, but how prominently.
The formula: average position = sum of your positions ÷ number of answers in which you appear. If a model lists you 1st, 3rd, 2nd and 6th across four answers, your average position is 3.0. Answers that do not mention you are left out, so average position should always be read next to mention rate.
The idea resembles average position in classic search reports, but AI answers have no fixed results page. Here, position means the order in which a brand appears in a generated list or text, and it can change every time the prompt runs.
Why Average Position matters
Readers tend to focus on the first recommendations in an answer. A brand named first is often framed as the default choice, while a brand named sixth can be skimmed past even though it technically counts as a mention. That is why positioning is one of the five factors in Trackbase's AI visibility score.
Averages can also hide shifts. A stable average of 3.0 can mean you are always third, or that you swing between first and last place. Track the distribution as well, for example the share of answers in which you are named first.
How to measure and improve Average Position
Use prompts that naturally produce lists, such as "best [category] for [audience]" or "top tools for [use case]".
Record your position in every answer for each model and compare it with competitors on the same prompts.
Look at the distribution, not only the average, and review it weekly.
Strengthen the signals that tend to lift brands higher in answers: consistent positioning across your site, reviews and comparison articles, clear proof points and positive third-party coverage.
Check the sentiment around your name. A first place with a caveat ("an older option") is worth less than a confident recommendation.
Trackbase Prompt Tracking measures where your brand appears in every answer across 11 AI models and shows how you rank against competitors. Why some brands end up on top is explained in how AI models decide which brands to recommend.
Examples of Average Position
A B2B payroll software vendor appears in 60 percent of comparison answers, but at an average position of 4.2, behind three competitors.
A Dutch bicycle webshop moves from an average position of 5 to 2 on e-bike prompts after publishing a comparison guide and earning coverage in a cycling magazine.
A project management tool holds a steady average of 2.5, yet its share of first places falls from 40 to 15 percent within a month.
A hotel group ranks first on Perplexity but fifth on ChatGPT for the same prompt, which points to differences in the sources each model uses.
Frequently asked questions
How is average position calculated in AI answers?
Add up your position in every answer where your brand appears and divide by the number of those answers. Being named first counts as 1, second as 2 and so on. Answers that do not mention you are excluded, which is why average position should always be reported together with mention rate. Calculate it per model and per prompt group.
Is a lower average position better?
Yes. Position 1 means you are the first brand named, so a lower number means more prominence. But a low average based on few answers can mislead: ranking first in two answers out of fifty says less than ranking second in forty. Combine position with mention rate and share of voice for a complete picture.
Why does my position differ between ChatGPT, Gemini and Perplexity?
Each assistant draws on its own mix of training data, retrieval methods and sources. It also builds its answer in its own way. A review site that one assistant relies on may carry little weight in another. These differences are useful: they show which platforms need attention and which sources to target for each of them.