How QuickSEO measures AI Visibility
Every metric you see in QuickSEO follows a published formula. The same formula is applied to your brand and to every tracked competitor — that's what makes the leaderboard comparable.
The four metrics
QuickSEO surfaces four numbers per brand. Each answers a different question.
| Metric | What it answers | Range |
|---|---|---|
| Visibility (Mention Rate) | How often is this brand named in AI answers? | 0–100% |
| Citation Rate | How often is this brand's domain cited as a source? | 0–100% |
| Share of Voice | What share of all tracked-brand mentions belong to this brand? | 0–100% |
| Sentiment (NSS) | When mentioned, is this brand framed positively or negatively? | −100 to +100 |
Visibility — the headline metric
Visibility = (prompts where brand is mentioned in the answer / total prompts) × 100
A brand is "mentioned" when its name (or any tracked alias) appears in the AI's answer text. We use an LLM to detect mentions — string matching alone misses paraphrases and pluralizations.
Worked example. You run 200 prompts × 4 platforms = 800 prompt runs in the last 30 days. The AI mentions your brand by name in 264 of those runs.
Visibility = (264 / 800) × 100 = 33.0%
The same formula is applied to every tracked competitor. If Salesforce is named in 432 of the 800 runs, Salesforce's Visibility is 54.0% — directly comparable to yours on the same denominator.
Why this is the headline. It's the most-used metric across the industry (Peec, Otterly, Profound, Trakkr all use this definition). It is absolute: adding or removing a competitor doesn't change your Visibility number. It moves only when the AI itself changes what it says about you.
Citation Rate — the source signal
Citation Rate = (prompts where the brand's domain appears as a source URL / total prompts) × 100
"Cited" is stricter than "mentioned". A brand is cited when one of its URLs (root domain or any subdomain) appears in the AI's source list for the answer — not just when its name is dropped in prose.
Worked example. Across the same 800 prompt runs, Salesforce.com appears in the source list of 144 runs.
Citation Rate = (144 / 800) × 100 = 18.0%
What it tells you. Citation Rate tracks whether AI is actually reading your content versus repeating reputational knowledge it already has. A brand can be highly mentioned without being cited (AI knows the name from training data) or highly cited without being mentioned (AI uses the content but credits a category instead of the brand). Both gaps suggest different fixes.
Citation Rate uses the same denominator as Visibility, so the two numbers sit on the same scale.
Share of Voice — the competitive lens
Share of Voice = (mentions_for_brand / sum of mentions across all tracked brands) × 100
Where Visibility is absolute, Share of Voice is relative. It answers: of the AI-mention pie shared by you and your tracked competitors, what slice belongs to you?
SoV is rendered as a dedicated column in the in-product leaderboard, immediately to the right of Visibility, so you can read absolute reach and competitive share on a single row.
Worked example. Across 800 prompt runs in the last 30 days, the LLM-detected mention counts are:
- You: 264 mentions
- Salesforce: 432 mentions
- HubSpot: 198 mentions
- Pipedrive: 106 mentions
Your SoV = (264 / (264 + 432 + 198 + 106)) × 100 = (264 / 1000) × 100 = 26.4%
Salesforce's SoV is 43.2%, HubSpot's is 19.8%, Pipedrive's is 10.6%. The four rows sum to 100% (small rounding deltas of ±0.1% are expected — each row is rounded to one decimal before display).
Watch out. Share of Voice depends entirely on which competitors you've tracked. Adding a heavyweight competitor drops everyone's SoV without anything actually changing in reality. We surface SoV alongside Visibility, not as a replacement — Visibility is the truer measure of how AI sees your brand. SoV is the marketing-friendly framing that's easier to digest in board reports.
Edge case. If none of the tracked brands are mentioned in any prompt run in the window, every row's SoV is 0% (we never divide by zero).
Sentiment — Net Sentiment Score
NSS = ((positive mentions − negative mentions) / total mentions) × 100
For every mention, the same LLM that detects the mention also classifies the sentiment toward the brand as positive, neutral, or negative. NSS combines those into a single number from −100 (universally negative) to +100 (universally positive).
Worked example. Of your 264 mentions:
- Positive: 178
- Neutral: 71
- Negative: 15
NSS = ((178 − 15) / 264) × 100 = 61.7
NSS is reported alongside Visibility, never folded into it. A brand can be highly visible and negatively framed — that's an actionable finding, and merging the two numbers would hide it.
Per-platform breakdown
Every metric is also broken down per platform (ChatGPT, Claude, Gemini, Perplexity, plus AI Overviews where supported). The rolled-up Visibility on your dashboard is computed from the union of platform runs; per-platform Visibility uses each platform's runs as its own denominator.
This matters because each platform reads the web differently. You might be cited 40% of the time by Perplexity (which heavily uses external sources) and only 8% of the time by ChatGPT (which leans more on training-time knowledge). Averaging hides the divergence.
How we count "mentioned" and "cited"
Both signals come from the same LLM parse call that QuickSEO runs against every AI answer:
- Mentioned — the LLM looks for the brand name, optionally extended by aliases you can configure on the site settings page. It catches simple paraphrases ("Salesforce" / "SFDC" / "salesforce.com") and the boundary cases where a brand is named but never linked.
- Cited — the AI platform itself returns a list of source URLs alongside the answer. We normalize each URL to its root domain (e.g.
help.salesforce.com→salesforce.com) and count one citation per (prompt run × domain) pair.
For competitors, the same LLM call processes every tracked competitor's name in a single pass — so the cost of adding more competitors to your tracking list is bounded, not linear.
Choosing prompts that produce stable numbers
These metrics get noisy on very small samples. As a rule of thumb:
- 20 or fewer prompts per platform per week: trend with care, single-week swings can be sampling noise.
- 50–100 prompts: stable enough for week-over-week tracking.
- 100+ prompts: stable enough for daily monitoring.
If your prompt count is below the comfort zone, QuickSEO surfaces a "low sample" warning on the dashboard.
Why not a single "AI Visibility Score" composite?
Many competitors ship a single proprietary 0–100 number that bakes mentions, citations, position, sentiment, and platform weighting into one figure. QuickSEO deliberately doesn't.
The trade-off is the same one Ahrefs Brand Radar made: a single composite is easy to read on a dashboard but impossible to defend. If your score moved from 64 to 71, what changed? Without the underlying inputs, you can't tell whether sentiment improved, citations grew, or one platform shifted weight.
QuickSEO surfaces the four atomic numbers and lets you drill into each. Visibility tells you reach; Citation Rate tells you source-of-truth; SoV tells you competitive share; Sentiment tells you framing. Together they answer questions a composite can't.