How Is an AI Visibility Score Calculated? Formula + Worked Example

The short answer

An AI visibility score measures how prominently a brand appears in AI-generated answers, on a 0–100 scale. It is calculated by asking AI answer engines a large set of non-branded, high-intent consumer questions, then scoring every answer on four weighted factors:

The weighted factors combine into a single score:

Visibility Score = (0.20 × Presence) + (0.25 × Position) + (0.25 × Coverage) + (0.30 × Share of Voice)

This is the exact formula behind the AI Visibility Index, which scores 600+ brands across 25 industries in India every week. The rest of this post unpacks each factor and walks through a real worked example.

There is no industry-standard formula — every tracking platform weighs things differently. What matters is that the methodology is published, the prompts are non-branded, and the measurement repeats on a fixed schedule so scores are comparable week over week.

Why non-branded prompts are the foundation

The single biggest methodological decision is what you ask the AI. A useful visibility score only uses non-branded prompts — the questions a real buyer asks before they know which brand they want:

• "Which bank is best for a savings account in India?"
• "Best phone under ₹30,000 right now"
• "Most reliable health insurance for a family of four"

If you ask branded questions ("Is HDFC Bank good?"), the brand obviously appears, and the score measures nothing. Non-branded prompts force every brand to compete for the same answer space — which is exactly what happens when a real customer asks.

Each category in the AI Visibility Index is tested with 25 non-branded, high-intent prompts, refreshed weekly against AI answer engines with live web search enabled. How to design that prompt set is a topic of its own — we covered it in how to choose the right prompts for AI search visibility.

The four factors, unpacked

1. Presence and mention quality — 20%

The baseline question: did the brand appear in the answer at all? But raw presence is too crude on its own, so mention quality matters too:

• Recommended — the answer explicitly suggests the brand ("HDFC Bank is a strong choice for...")
• Listed — the brand appears in a comparison or list without endorsement
• Mentioned in passing — named only as context ("unlike larger banks such as...")

A recommendation is worth substantially more than a passing mention. Sentiment is part of this factor as well — a brand named as a cautionary example is not "visible" in any useful sense.

2. Position in the answer — 25%

AI answers are read top-down, and answer engines themselves put their highest-confidence picks first. A brand named first in a "top 5" list captures most of the attention; the brand in fifth place captures a fraction of it.

Position scoring rewards being the first brand mentioned, decays for each subsequent slot, and is averaged across every answer the brand appears in. This is why two brands with identical mention counts can have very different scores — one is consistently the lead answer, the other is consistently the afterthought.

3. Query coverage — 25%

Coverage asks: across the full prompt set, in how many distinct questions does the brand show up?

A brand that appears in 20 of 25 prompts has broad relevance across the category. A brand that appears in 3 of 25 — even if it dominates those 3 — is a niche answer. Broad coverage is hard to fake, because it requires the AI to associate the brand with many different intents: best-for-beginners, best-value, most-premium, most-reliable, and so on.

4. Share of voice — 30%

The most heavily weighted factor, because it is inherently competitive. Share of voice is the brand's percentage of all brand mentions in the category. If AI answers across the prompt set mention banks 200 times and your bank accounts for 30 of those mentions, your share of voice is 15%.

Share of voice is a zero-sum metric: it can fall even when your mentions are stable, simply because a competitor started getting mentioned more. That makes it the best single indicator of competitive standing — and why it carries the largest weight.

Worked example: scoring a bank

Here is a simplified worked example using one brand in the banks category. Suppose across 25 weekly prompts the measurements come out as:

Applying the weights:

Score = (0.20 × 62) + (0.25 × 48) + (0.25 × 56) + (0.30 × 13 normalised vs category)
      = 12.4 + 12.0 + 14.0 + 3.9 (+ competitive normalisation)
      ≈ 42 / 100

A score of 42 sounds low — until you see the distribution. In real measured data, category leaders typically score in the 20–60 range, not 90+. No brand is mentioned in every answer, in first position, with majority share of voice. This is the most common misreading of visibility scores: they are relative competitive measures, not exam grades. You can see real distributions on any category page of the index, where bars are coloured relative to the category leader for exactly this reason.

When comparing scores, compare within a category and across weeks — not across categories. A 45 in banks and a 45 in skincare reflect different competitive densities and prompt sets.

What a weekly refresh adds

A single measurement is a snapshot; the value compounds when measurement repeats on a fixed schedule:

• Rank movement — which brands climbed or dropped since last week
• Trend direction — three consecutive weeks of falling share of voice is a signal, one week is noise
• New entrants — brands breaking into the top 25 for the first time
• Volatility — categories where AI answers churn weekly vs. categories with entrenched leaders

The AI Visibility Index refreshes every category weekly and publishes the movement openly — the full mechanics are documented on the methodology page.

How to improve each factor

Each factor responds to different work:

We maintain a practical playbook in AI visibility optimization: how to improve your score, and a deeper guide to writing AI-optimized content.

Frequently asked questions

question: How is an AI visibility score calculated? answer: By asking AI answer engines a fixed set of non-branded consumer questions, then scoring each brand on four weighted factors — presence and mention quality (20%), position in the answer (25%), query coverage (25%), and share of voice (30%) — combined into a 0–100 score.

question: What is a good AI visibility score? answer: In real measured data, category leaders typically score between 20 and 60, not 90+. Scores are relative competitive measures — compare your score against your category's leader and average, not against 100.

question: Why did my score change when nothing about my brand changed? answer: Share of voice is zero-sum — a competitor earning more mentions reduces yours. AI answers also shift as engines re-crawl the web, so weekly movement within a few points is normal; sustained multi-week trends are the meaningful signal.

question: Can I calculate my own AI visibility score manually? answer: At small scale, yes — ask 20–25 non-branded questions in an AI assistant, record which brands appear, in what order, and how often. The manual version misses weekly trends and competitive share of voice, which is what dedicated tracking automates.

question: Do all AI visibility tools calculate scores the same way? answer: No. There is no industry standard, and some platforms do not publish their formula at all. Before trusting a score, check that the methodology is public, the prompts are non-branded, and the measurement repeats on a fixed schedule.

This post was written by the PingAura Data Team, the team behind the LLM Visibility Index — tracking how brands rank in AI-generated answers across 25 major industries in India.