Metric Reference

One page to answer “what exactly is this number?”. Each row links to the methodology page that explains it in full.

Per-probe metrics

Computed for a single probe — one prompt, one engine, one region.

Metric	Raw range	Shown as	Applies to
Position	0–85	0–100 (raw ÷ 85 × 100)	Discovery prompts only; 0 for branded
Sentiment	0–15	0–100 (raw ÷ 15 × 100)	All intents
Frequency bonus	0–8	Folded into AI Score	Discovery prompts only
AI Score	0–100	0–100	All intents
Verdict	Enum	Badge	Branded prompts only
Mention status	Boolean	Mentioned / Missed	All intents

Verdict values: comparison prompts return win / tie / lose; trust prompts return legit / mixed / scam; discovery prompts return none.

Panel metrics

Averages over the probes in the current scope.

Metric	Formula	Denominator	Where
Final Score	0.8 × avg AI Score + 0.2 × SEO Score	Scorable probes	Headline card
AI Score	Mean of probe AI Scores	Scorable probes	Headline card
Sentiment	Mean sentiment × 100 ÷ 15	Scorable probes that mention the brand	Headline card
Mention Rate	Mentioned ÷ total	Scorable discovery probes	Headline card
Position	Mean position score	Scorable probes	Run statistics
Win rate	Wins ÷ comparison matchups	Comparison probes	Discovery tab
Visibility % (leaderboard)	Conversations naming a brand ÷ conversations	Discovery or branded bucket	Competitors tab

Site metrics

Metric	Range	Source	Note
SEO Score	0–100	Mean of four Lighthouse categories	Site-level — identical in every region
Performance	0–100	Lighthouse	Core Web Vitals
Accessibility	0–100	Lighthouse	Contrast, labels, keyboard navigation
Best Practices	0–100	Lighthouse	HTTPS, console errors, dependencies
SEO	0–100	Lighthouse	Crawlability, meta tags, mobile
Checks passed	Count	Structural GEO checks	Shown as passed of total

Counts on the panel

These three are not the same number, and the gaps between them are informative:

Count	Means
Probes executed	Prompt × engine × region calls the run made — this is what was deducted from your allowance
Probes averaged	Executed minus probes whose analysis failed
Mentioned	Scorable probes where the model found your brand

Exclusion rules at a glance

Rule	Applies to	Reason
Failed analyses excluded from everything	All metrics, numerator and denominator	Not measured ≠ measured badly
Sentiment counts mentioned probes only	Sentiment card and charts	An answer that never names you has no tone toward you
Mention Rate counts discovery only	Mention Rate, share of voice, competitor counts	Branded prompts guarantee a mention
Cross-region tables use geo-adjustable engines only	Region scorecard, regional charts	Otherwise regions have different denominators
Region average weights regions equally	Cross-region baseline	A timed-out probe shouldn’t shrink a market’s weight
SEO term dropped when no audit exists	Final Score	A missing audit is not a zero

Full reasoning in averages & denominators.