AI visibility audits for crawled pages
This page explains how GEO (Generative Engine Optimization) evaluates the latest crawl and exports a self-contained remediation package.
What is GEO?
GEO evaluates your crawled pages with a ranked hierarchy. Content quality matters most, then crawler access, then indexing and snippet eligibility, then structured data and entity context, then anti-patterns that suppress visibility.
Strong GEO signals help AI systems identify pages that are eligible, useful, and trustworthy enough to ground an answer. The goal is not special AI-only markup. The goal is original, crawlable, technically eligible content.
How the scorecard works
Open GEO from the site's left navigation, or select GEO in the Knowledge base header. The GEO header links back to Knowledge base.
After a completed crawl, AI visibility scorecard scores five layers from 0 to 3 for a total out of 15.
| Layer | Meaning |
|---|---|
| Content quality | Checks whether pages contain original evidence, first-hand experience, expert attribution, proprietary examples, or unique point of view. |
| Crawler access | Checks whether robots.txt allows search and AI retrieval crawlers needed for citation visibility. |
| Indexing and snippets | Checks canonical tags, noindex/nosnippet directives, titles, descriptions, headings, mobile viewport, and crawlable HTML. |
| Structured data | Checks JSON-LD entity types and Open Graph metadata where they support page understanding. |
| Anti-patterns | Flags duplicated pages, thin templated sections, keyword-variation patterns, and other low-value tactics. |
The Highest leverage fixes list is ranked by impact. Selecting a fix scrolls to the export section. If crawl data exists but the hierarchy audit has not been generated, the page shows Audit scorecard pending and asks for a fresh crawl.
What do findings mean?
Findings are page-level evidence behind the scorecard. They are organized by severity:
| Severity | Meaning |
|---|---|
| Critical | A blocker that can disqualify pages from indexing, snippets, or crawler access. |
| Warning | A material weakness that limits AI citation readiness. |
| Info | A lower-risk improvement opportunity. |
The Crawl overview reports pages, all findings, and critical, warning, and info counts for the latest crawl. Treat critical findings first, but do not mistake technical cleanup for strategy. If content quality is weak, the highest-leverage fix is usually original examples, real data, named expertise, and first-hand perspective.
How do coding agent exports work?
When crawl data exists, Coding agent export builds two client-side files from the current scorecard, pages, and findings:
- Remediation JSON is a structured, self-contained task package.
- Remediation brief is markdown intended to be read by a person or pasted into a coding agent.
Use Copy JSON, Download JSON, Copy markdown, or Download .md. The status banner confirms a copy or names the downloaded file. Switch the preview between Remediation JSON and Remediation brief, or use Hide preview when you only need the download actions.
Exports do not apply changes to your website and do not send tasks to an external agent. Download or copy a package, give it to your preferred coding agent, review the resulting changes, publish them, and run a fresh crawl to measure the result.
What happens without crawl data?
GEO evaluates the latest crawl job for the selected site. Before the first crawl, the page shows No crawl data yet and directs you to Crawl. A running or failed crawl is reflected by the status badge; complete or retry that crawl before treating the findings as current.