Metric definitions
Public definitions of how scores and statistics are computed.
AI visibility score
Visibility = mention rate 40% + position bucket 25% + citation coverage 20% + cross-engine consistency 15%. Citation coverage only counts engines that return structured citations; when a snapshot has none, the remaining dimensions are renormalized to sum to 100. Each score is computed per snapshot and averaged across engines.
Mentions vs citations
A mention names your brand in the answer; a citation is a linked source. Tracked separately because being named is worth more than being linked.
Position bucket
Where your brand first appears in the answer, scored 1–10: the earlier the mention, the higher the score, and a mention in the first sentence scores highest.
Cross-engine consistency
How stable your presence is across engines for the same keyword. High variance means the signal is unstable.
Share of voice
Brand mention count ÷ total mentions of the brand plus all co-listed competitor brands. Measures your share of the conversation in those answers. Not computed when the brand is never mentioned.
Co-listed ranking
Ranks your brand against every other brand in the same answers by mention count, and shows your position (N of M) — a quick read on where you stand.
Citation gaps
Domains AI cites in answers that mention your brand, excluding your own domain (when your brand isn't mentioned, all answers count). Use it to prioritize the channels where citations need to be built.
Top cited domains
The most-cited source domains for each keyword — the channels and content entry points that matter most.
Brand domain derivation
We infer your brand domain from the company name (e.g. Acme Valves → acme-valves.com); if the name is already a domain, we use it as-is. You can set the domain manually per project, and we'll use that instead.
95% confidence intervals
Mention and citation rates use the Wilson interval (stable at small sample sizes); the overall score uses deterministic bootstrap resampling (400 iterations). Shown on Professional reports and keyword tables to convey statistical confidence given snapshot volume.
Negative signals & semantic roles
Professional adds entity-level analysis: when an answer names your brand alongside negative or risk language, the signal is flagged and counted as a keyword warning; a statement with no source attached is flagged as an uncited claim. Both drive content and citation priorities.
Action-plan evidence
Every recommendation in the action plan includes the evidence that triggered it (keyword mention rates, citation-gap domains, keywords captured by competitors, etc.), so each action is traceable and reviewable.
Site audit
The site audit checks four signal groups: crawlability (HTTPS active, homepage reachable, robots.txt allows AI crawlers, sitemap.xml, llms.txt), structured data (Organization/WebSite/Product/FAQPage), citation signals (industry directories, marketplace profiles and other citable sources) and content quality (factual specs, customer FAQs), producing a 0–100 citability score.
Snapshots & re-measure
Each audit produces a snapshot, retained by the plan at creation time: 1 month on Free, 12 months on Lite/Standard/Professional; changing plans never deletes existing snapshots. Deleting a project cascades all snapshots. Re-run after changes to compare the same keywords.
Credit basis
Each engine query consumes 1 credit; one audit consumes keywords × prompts × engines. Prompt count is the number actually selected — at least 1 per keyword (up to 5 on Free). Free includes 20 one-time credits (no monthly reset); Lite 200/month; Standard 500/month; Professional 2,000/month. Audits pause when credits run out and prompt you to view plans or add credits.
Demo data
Demo projects use fixed seed data labeled everywhere as demo. Failed engine calls mark the audit DEGRADED with error details, never silent fallback.