Learn how to read your GEO audit scores, from the AI Visibility Index down to individual sub-scores and benchmarks.
Last updated
Your audit report breaks AI visibility into layers of scores. Each layer measures a different aspect of your page's readiness for AI citation. This page explains what every score means, how it is calculated, and what targets to aim for.
For a high-level overview of the scoring framework, see Metrics Explained.
The AI Visibility Index (AIVI) is your headline metric. It combines optimization readiness with actual AI platform performance into a single number.
(GEO Readiness × 0.6) + (Share of Voice × 0.4)The 60/40 weighting reflects that optimization is what you control directly. Share of Voice — how often AI platforms mention you — tends to lag behind the improvements you make.
GEO Readiness measures how well your content is structured for AI citation. It is a dynamic average of your pillar scores.
When you run a standard audit, three pillars contribute: Technical, Content, and Authority. When you enable media analysis, Media becomes the fourth pillar and the average adjusts automatically.
The report lists each pillar's score under the headline, after "GEO Readiness is the average of". Select a pillar to jump to its section.
If an audit could not read your page, Content is not measured and counts as 0 in GEO Readiness. The report and the audit preview say so beside the pillar scores. To let the audit in, see Let Vizzybl read your site.
Share of Voice is the measured half of the index: how often AI engines mention your brand.
During an audit, Vizzybl asks a small set of questions about your market to each AI engine in your plan, then checks every answer for your brand.
This is a snapshot taken at audit time. The Share of Voice on your dashboard comes from prompt monitoring, which tracks your own prompts continuously.
The Technical pillar evaluates whether AI engines can access and parse your content.
The Technical pillar starts at 100 and loses points for each issue the audit finds. In the report's Infrastructure & Crawlability section, the What Cost Points card lists each issue with the points it cost, largest first, so the list adds up to your Technical score. Anything else the audit noted that cost nothing appears under Also noted.
| Issue | Points lost |
|---|---|
AI crawlers blocked in robots.txt | Up to 30 |
robots.txt names no AI crawler | 5 |
No llms.txt file | 10 |
| No schema.org markup | 20 |
Page has a noindex directive | 25 |
| No canonical link | 3 |
| Canonical link points to a different page | 5 |
| Page not served over HTTPS | 10 |
| No mobile viewport tag | 5 |
A blocked crawler that fetches pages for live answers costs more than one that collects training data. Blocked and unnamed crawlers together never cost more than 30. On older audits where the points cannot be added up exactly, the card lists the findings without points.
Semantic HTML measures how well your page uses structured markup. It is the Content pillar's structure score, not a Technical input — the Technical pillar covers AI bot access, llms.txt presence, schema presence, and indexability. Three sub-scores contribute to the total.
| Sub-Score | Max Points | What it measures |
|---|---|---|
| Heading Hierarchy | 40 | Valid outline with no skipped levels and a single H1 |
| Semantic Containers | 30 | Presence of main, article, nav, aside, header, footer |
| Clean Scope | 30 | Main content ratio vs. noisy sidebar and footer elements |
Vizzybl checks for structured data types that AI engines use to extract facts: FAQPage, Product, HowTo, and Article. Pages with valid schema markup are flagged as rich-result eligible.
Your robots.txt configuration determines which AI crawlers can read your pages. The Server Access Checker card lists every AI crawler the audit checks, whether it is allowed or blocked, and whether robots.txt names it or it relies on a default rule.
Blocked crawlers reduce your Technical score. So does a robots.txt that names no AI crawler at all: several answer engines treat a default rule as ambiguous, so the report rates this Needs improvement even when nothing is blocked. Add an explicit Allow rule for each AI crawler you want to reach you. Vizzybl also checks whether an llms.txt file is present.
Two cards in the report are shown for reference. Neither is part of any score.
The Content pillar measures the quality and structure of your text for AI extraction. Four inputs contribute: nugget density, structure score, readability, and freshness. Transparency is also measured during the audit, but it feeds the Authority pillar rather than this one.
Freshness has the strongest correlation (r=0.68) with AI discoverability in the GEO-16 framework. Three sub-scores contribute.
| Sub-Score | Max Points | What it measures |
|---|---|---|
| Date Presence | 30 | Schema.org dates, meta tag dates, and visible dates on the page |
| Recency | 40 | How recently the content was last updated |
| Human Visibility | 30 | Whether dates are visible to readers, not hidden in metadata |
Content updated within the last 30 days earns the maximum recency score. Content older than two years earns the minimum.
Transparency measures trust signals that AI models look for when deciding whether to cite a source.
| Sub-Score | Max Points | What it measures |
|---|---|---|
| Author Attribution | 30 | Named author, schema markup, credentials, and linked profiles |
| Editorial Disclosure | 30 | Disclaimers, editor's notes, and sponsored content notices |
| Provenance | 40 | Evidence-based claims, source attributions, and methodology sections |
The Authority pillar measures your brand's recognition and trustworthiness.
Vizzybl queries the Google Knowledge Graph API to determine whether your brand is a recognized entity. The result score maps to trust levels.
| Result Score | Trust Level |
|---|---|
| Above 700 | High — Strong entity recognition |
| 400–700 | Moderate — Partial recognition |
| Below 400 | Low — Weak or no entity presence |
Vizzybl compares your brand description with how the Knowledge Graph describes you. The result is classified as match, partial_match, mismatch, or no_kg_description. Closer alignment signals stronger brand consistency to AI engines.
The Media pillar is optional and appears when you enable media analysis during an audit. It evaluates how well your visual content supports AI extraction.
Each image is scored across four dimensions.
| Dimension | Max Points | What it measures |
|---|---|---|
| Contextual Metadata | 25 | Alt text quality, filename descriptiveness, caption alignment |
| Visual Comprehension | 30 | Entity recognition, OCR readability, topic alignment |
| Technical Discoverability | 25 | ImageObject schema, file format, size optimization |
| Originality | 20 | Custom vs. stock indicators, branding, metadata richness |
Videos are evaluated for timestamp and chapter presence, transcript availability, VideoObject schema completeness, and answer-first structure.
The Citation Potential Score (CPS) predicts how likely AI engines are to cite your page. It combines four weighted factors.
| Component | Weight | What it measures |
|---|---|---|
| Fact Density | 40% | Statistics, percentages, data points, and verifiable claims |
| Novelty | 35% | How unique your content is compared to existing search results |
| Link Authority | 15% | Quality of outbound links to authoritative domains |
| Data Formatting | 10% | Presence of tables, lists, and structured data |
Formula: (0.40 × factDensity) + (0.35 × noveltyScore) + (0.15 × linkAuthority) + (0.10 × dataFormatting)
Novelty comes from comparing your page with the search results for your keywords. When that comparison could not be made, the report shows Not measured in this audit for Novelty in place of a score.
Use these benchmarks to interpret any score in your report.
| Score | Rating | What to do |
|---|---|---|
| 80–100 | Excellent | Maintain and monitor for regressions |
| 60–79 | Good | Address specific gaps in your weakest pillar |
| 40–59 | Needs work | Prioritize your lowest-scoring pillar for improvement |
| 0–39 | Critical | Fundamental optimization needed across multiple areas |