OVERVIEW

The GEO Auditor measures whether a website is easy for modern AI systems to access, interpret, and cite. It is built for teams who need a practical view of AI readiness, not just classic search visibility.

In the web product, audits run in the background, stream progress live to the browser, and finish with stakeholder-friendly outputs such as an Excel workbook, optional PDF, topic map, heatmap, schema recommendations, and optional llms.txt generation.

WHAT THE SYSTEM CHECKS

The audit combines site-wide, page-level, and crawl-level signals:

• Robots and AI Access: Checks crawler rules in robots.txt and evaluates whether an existing llms.txt is missing, thin, weak, or useful.
• Schema, Metadata, and HTML Structure: Measures how clearly the page explains what it is, what it covers, and how the content is organized.
• Content Quality: Looks at depth, formatting, and semantic coherence so the page is not just indexable, but also quotable.
• Entities: Checks whether brands, people, places, and products can be extracted cleanly and mapped to canonical references such as Wikidata.
• Chunk Citability: Scores whether individual passages are self-contained and useful enough to be lifted into an AI answer.
• Authority: Combines domain age, entity schema strength, external brand signals, grounding-page signals, and optional backlink proxy data.
• Agentic Readiness: Looks for practical signals such as WebMCP annotations and a UCP discovery file at /.well-known/ucp.
• Harmonic Centrality: In crawl mode, the system checks which pages are central or peripheral in the discovered internal link graph.
• Rendering Consistency: When both raw and JS-rendered HTML are collected, the audit shows whether crawlers and users are effectively seeing the same content.

HOW A WEB AUDIT RUNS

• 1. URL Resolution: The system resolves the audit set from sitemap discovery, scoped crawl mode, exact uploaded URLs, or uploaded seed URLs.
• 2. Site-wide Checks: Domain-level signals such as robots and authority run once and are reused across the pages in the same job.
• 3. Background Page Audits: Each page is processed by the worker queue while the browser receives live progress updates.
• 4. Finalization: The system aggregates scores, prepares factor summaries, and writes the final report artifacts.

INPUT MODES

• Direct URL: Start from a single site or page in the main input field.
• Exact URLs: Upload a file and audit only the listed URLs.
• Seed Crawl: Upload one or more starting URLs and let the crawler discover additional pages from there.
• Scope Controls: Use max URLs, JS sample size, optional path scope, and optional locale filters to shape the run.

OUTPUTS

• Excel Workbook: The main technical export for the job.
• PDF Report: Optional stakeholder-facing summary.
• schemas.json: Recommended schema templates based on the audited domain.
• Topic Map: An HTML graph showing page and entity relationships.
• Chunk Heatmap: A visual pass/fail style view of citability by page section.
• llms.txt: Optionally generated only when requested and only when the audited site does not already expose one.

HOW TO READ RESULTS

• Weighted factors move the headline score.
• Supporting factors add context but do not change the headline score.
• Diagnostic factors explain technical risk.
• Not measured does not automatically mean failure. It can also mean the signal was intentionally unavailable for that run.