AI Visibility Score Explained: What 0-100 Really Means

The AI Visibility Score measures your brand's visibility across AI models on a 0 to 100 scale. But what does that number actually mean? More importantly, what can you do to change it? This guide breaks down the four weighted components, the score bands, and what a single fix in each component typically delivers.

The four components and why they're weighted this way

The TopSlot Visibility Score is the weighted combination of four signals. Each measures a different facet of appearing in an AI answer, and the weighting reflects how much each one drives buyer behavior in practice.

Mentions — 40% of the score

Mentions answers the gate question: does your brand name appear in the AI's response to a relevant buyer query? Nothing else matters if the answer is no. We weight mentions highest because absence beats every other signal in impact — if you're not in the answer, no amount of prominence or sentiment optimization rescues you.

A mention can take three forms in practice: cited with URL (the model links to your domain), named only (your brand verbatim, no link), or descriptive only (the model describes what you do without naming you, detected via embedding similarity). All three count as mentions in this component, but they're weighted differently downstream by the Citations component.

Prominence — 30% of the score

Prominence captures where in the answer your brand appears. First sentence vs. seventh in a list is a 5x difference in buyer recall. Top-three vs. tail-of-list is roughly 3x. Prominence is how we encode the AI's implicit ranking — the model is telling you something when it puts a competitor before you in the same answer.

We measure prominence as a positional rank, normalized for the length of the answer. A brand named first in a five-brand list scores higher prominence than a brand named first in a fifteen-brand list, because the second case is essentially "we mentioned everyone in this category".

Citations — 20% of the score

Citations specifically tracks whether the AI links back to your URL versus just naming you. A cited URL drives traffic, builds verifiable authority, and reinforces your domain in the next training run. Naming-only is valuable for buyer shortlist construction but doesn't deliver the downstream metrics.

Citation rates vary enormously by model. Perplexity cites almost every answer with numbered footnotes. Gemini cites via source chips. ChatGPT cites only when browsing is invoked; pure-training responses don't cite at all. Claude cites when its web search tool fires. This means your Citations score is partly a function of which models your buyers actually use.

Sentiment — 10% of the score

Sentiment measures whether the AI describes you positively ("leader", "recommended"), neutrally, or with negative framing. We weight it lowest because it's fragile — a single negative review can swing per-model sentiment — and because the other three components do most of the work to drive buyer behavior. But you ignore sentiment at your peril if you're in a category where trust is the primary purchase driver (financial services, healthcare, security).

Score bands and what they mean in practice

From 200+ scorecards we've run across categories: 0–10 means AI does not know you exist. Your brand never appears in category-level buyer queries. Fixes are entity-existence basics: get into Wikipedia if eligible, build a real LinkedIn company page, get listed on the obvious review sites. 11–30 is emerging visibility. You appear in some niche queries but not the broad category ones. 31–50 is moderate. Buyers see you in roughly half of relevant queries. 51–70 is strong — you're a default candidate in most category answers. 71–85 is dominant. You're named first or near-first in nearly every relevant answer. Above 85 is rare and usually reserved for category-defining brands.

A single score is less useful than the trajectory. A brand at 45 dropping is in a worse position than a brand at 30 climbing. Track the slope over weeks, not the snapshot. The AI Strategy Advisor builds this trend view directly into the weekly Wins/Alerts/Opportunities briefing.

Why per-model breakdown matters more than the composite

A brand at 45 composite with a per-model split of 80/60/30/10 across ChatGPT/Gemini/Claude/Perplexity has a very different problem than a brand at 45 with a flat 45/45/45/45. The first has a Perplexity-specific freshness gap — Perplexity favors content updated within 90 days, so refreshing your top pages will move that single number sharply. The second has a base-authority gap that affects every model uniformly — the fix is multi-quarter third-party validation work.

The diagnostic is in the variance. High variance = model-specific tactical fix. Low variance = structural authority work.

What moves each component

To move Mentions: create comparison pages, get on review sites, build Reddit presence, allow GPTBot/PerplexityBot/ClaudeBot in robots.txt, ensure SSR HTML works (these crawlers don't execute JS reliably). To move Prominence: earn third-party authority that reinforces your category leadership; comparison pages that explicitly position you above competitors; original data citations. To move Citations: publish content with extractable statistics, expert quotes, and structured data; keep dateModified fresh for Perplexity; add llms.txt. To move Sentiment: address any common complaint themes visible on Reddit and review sites; build a body of positive third-party coverage that outweighs occasional negatives.

How TopSlot calculates and tracks your score

The free AI Visibility Scorecard runs a single audit across two models in 60 seconds and returns a category-level result. The paid AI Search Tracker runs up to 25 buyer-intent prompts four times a day across all four models, generating a daily score with full per-model breakdowns, prompt-level diagnostics, and competitor comparison. The AI Ranking module extends this to your real Search Console queries — so the score reflects your actual buyer language, not a generic prompt list.

Start with the free scorecard. It tells you which band you're in and which component is your biggest gap. Stack fixes from there.