What every metric means
Every number Bizaura shows, defined in plain language, including the parts that make the numbers smaller than you might like. No metric here is a black box.
The audit
GEO Score
0–100 · how ready your site is to be cited by AI engines
A weighted composite of five measured signals, anchored in published GEO research (see the method post for the full story):
| SIGNAL | WEIGHT | WHAT IT MEASURES |
|---|---|---|
| Content quality | 25% | answer-first structure + fact density, one AI pass |
| AI crawlability | 20% | robots.txt access for 6 AI crawlers; all blocked = hard cap |
| Structured data | 20% | 8 GEO-relevant schema.org types found as valid JSON-LD |
| Freshness | 20% | dateModified / “last updated” signals; none anywhere = capped |
| Semantic HTML | 15% | h1/h2 hierarchy, no skipped levels, deterministic |
| Performance | 0% | PageSpeed / LCP shown for context only, never scored |
Two hard gates keep it honest: blocking all six AI crawlers caps the total score regardless of everything else, and having no date signals anywhere caps freshness. The audit reads at most 9 pages with a plain HTTP fetch (no headless browser), roughly what AI crawlers themselves see.
Monitoring
Monitoring asks the real AI assistants the questions your buyers ask (in their language), archives every verbatim answer, and measures who got recommended. Answers are collected by scraping the live ChatGPT / Gemini interfaces with web search, plus Google's AI Overviews from real result pages. They may still vary from your personal view due to personalization.
Visibility
% of tracked AI answers that mention your brand at all
Out of all answers collected in a run (every prompt × every engine), the share that names your brand at least once. 50% visibility = you appeared in half the answers a shopper could have gotten. Honest denominator: when an engine shows nothing for a prompt (for example, Google displays no AI Overview), that still counts as an answer you weren't in, because the shopper still didn't see you.
Share of Voice
your slice of all brand mentions across a run's answers
Your mentions divided by all brand mentions in the run, counting every brand the AI actually named, not just the competitors you configured. That makes it a real market share of AI attention: if the engines name 60 brands and you're 3 of 100 total mentions, your SoV is 3%. The weighted variant counts a mention more when it appears earlier in the answer (weight 1/position), since shoppers rarely read past the first few recommendations.
Position
where in the answer you appear, 1 means named first
The order of first appearance in each answer that mentions you: 1 = the first brand named, 2 = second, and so on. We report the average across answers and the best single placement. Lower is better: being brand #7 in a list of 8 is technically visibility, but it rarely wins the sale.
Sentiment
how the answer talks about you: positive / neutral / negative
Each mention is classified as positive (recommended, praised), neutral (listed without judgment) or negative (criticized, warned about), and tallied per brand. Most retail mentions are neutral-to-positive; a negative streak is an early warning that engines are citing complaints.
Cited sources
the pages AI answers actually linked to
The URLs and domains the engines cited while answering: your pages, competitors', retailers', comparison sites. This is the map of who feeds the answers: if a comparison site is cited in every answer and you're not on it, that's an actionable gap.
Reading the numbers together
The audit score says whether engines can cite you; monitoring says whether they do. The intended loop: run the audit, fix what it flags, then watch Visibility and Share of Voice trend in monitoring over the following weeks. One run is a snapshot; the trend across runs is the real signal.
Who else processes this data
Two services see data on our behalf. DataForSEO collects the engine answers: it runs your prompts against the live assistants and returns what they said. logo.dev serves the brand logos you see next to companies and cited sources in the dashboard: it receives the domains we request logos for, and nothing else. Those domains are business entities, your competitors and the sites the engines cite, not personal data, but we list the service here because it is a third party that receives something.
Logos provided by Logo.dev.