AI brand monitoring: Share of Voice and brand visibility in AI answers

The new metric family

A classic tracker measures position. GEO metrics measure presence in the answer:

• AI Citation Frequency (AICF) — how often AI cites you.
• Brand Representation Accuracy — how accurately the engine describes your brand.
• AI Share of Voice — your share of answers versus competitors.
• AI Referral Traffic — real people arriving from AI answers.

The measurement queries are grouped by funnel — 5 categories: commercial + agentic, navigational, comparative, recommendational, reputational. The KPI shifts from rankings to sales and presence (GEO-HowTo 2026, Runov).

AISVS: a 0–100 composite

One well-developed approach is AISVS (AI Search Visibility Score) by Dmitry Ivanov (ISPOLIN). It’s a 0–100 score from four independent signals:

Component	Weight	What it measures
Citation Rate	35%	share of queries where the brand appeared at all
Position Score	25%	where in the answer you were mentioned
Answer Share	25%	share of brand words versus all competitors
Persistence	15%	how many weeks in a row the brand holds

A few important details:

• Position Score splits the answer into zones: the start (first 80 words) weighs 3.4, the middle 1.0, the tail 0.4. Being mentioned in the first paragraph ≠ being mentioned at the end.
• Answer Share compares you against competitors found automatically: domains appearing 2+ times in answers, minus noise (Wikipedia, YouTube).
• Persistence is the time dimension. “Landing in Alice once is easy; holding for 26 weeks is the task.”

How to read it: 0–20 is a blind zone, 50 is baseline visibility, 70+ is strong. A telling ISPOLIN example: a company with AISVS = 28 while its overall technical score was 71/100 — a good base, but the brand hasn’t taken hold in AI answers yet. That’s exactly the gap only a GEO metric sees.

Sentiment: not just “in,” but “how described”

Making it into the answer isn’t enough — what the AI says about you matters. So the AI SOV model (Rusakov) adds a mention-sentiment weight:

Sentiment	Weight
Positive	1.5
Neutral	1.0
Negative	0.5

Different platforms count their own way: Ivanov’s AISVS emphasizes stability (Persistence), Rusakov’s AI SOV emphasizes factor weights (platform, link, sentiment), Brandfound’s BMR emphasizes before/after delta across 9 neural nets. There’s no single industry standard yet — but the direction is shared: measure presence in the answer, not the link’s position.

How we measure it

Our visibility report brings these signals together:

• AI-visibility score — a composite visibility score (AI-SOV, AI-Overview exposure, citation trend, page readiness) — our analogue of AISVS.
• Brand-perception with sentiment — the engine is asked “what is <brand>” and we read whether it knows you, describes you accurately, with what sentiment (positive/neutral/negative), and whether it cites your site as the source.
• Per-query competitors — for each query you see who’s cited instead of you (those 2+-times domains).
• AI-referrals — real visits from AI answers, per engine.

The industry’s baseline measurement basket is ~20 prompts × 3 engines, twice a month (ISPOLIN’s runs ~45 ₽ each). Regularity beats a one-off snapshot: it’s what produces Persistence.

What to do

1. Replace the headline metric. Position → share of citations (AI SOV / AISVS). Otherwise visibility growth is invisible.
2. Measure regularly. Twice a month against a fixed prompt basket — that’s how Persistence appears.
3. Watch sentiment. Separately from the fact of citation: negative in an answer costs more than silence.
4. Split prompts by funnel. 5 categories (commercial/navigational/comparative/recommendational/reputational) — otherwise the metric measures only part of demand.

Sources

• AISVS — AI Search Visibility Score — GEO-HowTo 2026, talk by Dmitry Ivanov (ISPOLIN): components Citation Rate 35% / Position Score 25% / Answer Share 25% / Persistence 15%, a ~20-prompt × 3-engine basket; methodological basis — Digital Applied.
• AI Share of Voice + sentiment weights — GEO-HowTo 2026, talk by Rusakov–Sinitsyn (Sentiment 1.5/1.0/0.5, PlatformWeight, LinkWeight); Brandfound BMR (delta across 9 neural nets).
• The GEO metric family and the 5 prompt categories — GEO-HowTo 2026, talk by Alexander Runov (MGCom).
• New visibility metrics (AICF, brand accuracy, SOV) — Digital Agency Network: GEO Statistics 2026, SEO.com: GEO trends.