The AIVI score: How Sparkloft measures AI visibility from 0 to 100
The AIVI score is a 0 to 100 index of how well AI platforms put a brand in front of people who ask for options without naming it. It blends four pillars, calculated on unbranded questions only, and is reported overall, by AI platform and by topic.
The four pillars
| Pillar | What it measures |
|---|---|
| Mention | Share of responses to unbranded questions that name the brand |
| Citation | Share of responses to unbranded questions that cite the brand's own domain as a source |
| Position | 100 divided by the brand's average rank among the options named, on responses that mention it |
| Sentiment | Average tone of the brand's mentions, classified by analysts against a ten-type tone taxonomy |
Why the score uses unbranded questions only
Questions that name a brand test recall of that brand. Questions that do not name it test whether AI recommends it. AIVI scores the second kind: discovery questions with no brand or place named, persona-shaped questions built around a person and situation, and category questions built around an interest, sector or need. Brand-named and comparison questions are reported separately.
How we keep the score reliable
AI answers vary from one run to the next, so a single response is an anecdote. AIVI runs every prompt three times on each platform, freezes the panel for each measurement period, and reports confidence ranges. Small subgroups are flagged as directional.
Keep reading
Start with a baseline
No existing baseline is needed; establishing one is the first deliverable of AI Visibility Intelligence (AIVI).
