Skip to main content
Trackbase turns every AI answer into a small set of metrics. Each one answers a different question. This page gives the short version of each and explains how the dashboard numbers are built. The Trackbase dashboard with cards for share of voice, visibility, presence rate and average position above the visibility over time chart. Example data.

​
From one answer to a score

Every answer a model gives to one of your prompts is analyzed for:
  • Brand mention: whether your brand name or one of its aliases appears, and how often. Matching is on whole words, so a brand name is not counted when it is only part of another word.
  • Position: where your brand first appears in the answer, counted from 1.
  • Sentiment: how positively the answer describes your brand, from 0 to 100. Sentiment is only scored when your brand is mentioned.
  • Sources: the URLs the model cited.
  • Competitors: which competitors are mentioned, and how often.

​
The metrics

​
How period numbers are calculated

  • The daily visibility score is the average over all answers of that day, including the answers that do not mention you.
  • If a prompt runs more than once on the same model on one day, only the first run counts, so daily numbers stay stable.
  • A metric card shows the average over the selected period. Mentions and citations show the total.
  • The change shown on a card compares the second half of the selected period with the first half.
  • The trend chart shows a 7-day moving average and leaves out the current day, because that day is still collecting answers.

​
How to read them together

  • Low presence rate, good average position: when you appear you are placed well, but you are missing from many answers. The work is coverage.
  • High presence rate, late average position: you appear often, but after your competitors. The work is prominence.
  • Mentions without citations: models know your brand from other sources, but your own pages are not feeding the answers. The work is content and crawlability.
  • Good visibility, weak sentiment: you are visible but described negatively. Read the answers to see which claims cause it.
When several metrics fall on the same day, check first whether you added prompts or models. New prompts where you are not mentioned yet lower the average without any change in how AI sees you.

​
Read more per metric

For definitions of the terms used here, see the glossary.
Last modified on November 12, 2026