Open Methodology
Audit Methodology v2.1
This page publishes the site-audit categories, weights, score definitions, change log, and scope. It is not the measurement infrastructure or dataset: fixed panels, entity normalization, parsing/error handling, score calibration, abuse controls, benchmarks, and client retest records stay private.
Scoring Bands
Each check scores full / half / zero points for pass / warn / fail. Layer-3 probe results are independent of this score — they measure, not diagnose.
PUBLIC
Auditable site standards
Categories and weights, check meanings, score definitions, version history, applicability and limits. A score does not predict traffic, revenue, or engine rank.
PRIVATE
Compounding measurement assets
Fixed query panels, model parameters, entity normalization, raw evidence, calibration, client change-to-retest chains, and benchmarks. Transparency does not mean opening measurement assets.
Crawlability 25
Whether AI crawlers can physically reach and read your page — robots rules, edge blocking, rendering, speed, transport.
Robots.txt AI Bot Rules
6 ptsEnsure robots.txt does not disallow major AI crawlers from accessing your site.
SourceRFC 9309 (Robots Exclusion Protocol); each vendor's official crawler documentation.
AI Crawler Reachability at the Edge
5 ptsDetect whether a CDN or WAF silently blocks AI crawlers at the edge.
SourceCloudflare AI Crawl Control / managed bot blocking documentation (2025-2026); measured behavior of managed bot rules.
Server-Side Rendering
6 ptsEnsure the page has visible content even without JavaScript execution.
SourceOfficial statements from OpenAI/Anthropic/Perplexity crawler docs: HTML is fetched without JS execution.
Page Response Speed
4 ptsKeep server response fast enough not to hurt crawl budget.
SourceStandard crawl-budget behavior; consistent with Google crawl scheduling guidance.
HTTPS
4 ptsThe page must be served over HTTPS with a valid certificate.
SourceBaseline transport security.
Content Structure 30
Whether the page is structured so an AI model can extract answers — hierarchy, Q&A blocks, metadata, and citable content signals.
Heading Hierarchy
6 ptsUse clear heading structure so AI models can segment content into quotable units.
SourceVendor guidance on structured content for AI extraction (Anthropic "Writing for Agents", OpenAI docs).
FAQ / Q&A Content
4 ptsInclude FAQ or Q&A-style content — the format AI answers quote most readily.
SourceQ&A blocks are the unit AI answers quote most readily; consistent with schema.org FAQPage intended use.
Meta Description
5 ptsProvide a meta description of appropriate length as a snippet candidate.
SourceSearch rendering window conventions.
Open Graph Tags
4 ptsInclude Open Graph tags for structured social sharing.
SourceOpen Graph protocol specification.
Direct Answer Blocks
6 ptsFollow headings with self-contained answer paragraphs AI can lift verbatim.
SourceFeatured-snippet extraction patterns, measured across the GEO audit corpus.
Content Citability (Quotes & Statistics)
5 ptsInclude statistics and direct quotations to raise citation probability.
SourceAggarwal et al., "GEO: Generative Engine Optimization", KDD 2024 (arXiv:2311.09735): cite-sources/quotation/statistics strategies raised AI visibility 30-40%; keyword stuffing was the only negative strategy. Note: C-SEO Bench (NeurIPS 2025) failed to reproduce part of the lift — treat as guidance and validate with your own probe results.
Authority Signals 22
Signals that make a page worth citing — machine-readable entity data, external references, authorship, and index coverage.
Schema Markup
7 ptsUse Schema structured data so AI can parse page content.
Sourceschema.org specification.
Schema Type Coverage
5 ptsCover multiple AI-friendly Schema types.
Sourceschema.org type usage in AI-cited pages.
External Citations
5 ptsCite authoritative external sources to build credibility.
Source"Cite Sources" strategy in the KDD 2024 GEO study (largest measured lift, up to 115% for low-visibility pages).
Author Attribution
3 ptsShow clear author attribution for content trustworthiness.
SourceE-E-A-T style attribution signals.
LLMs.txt File
2 ptsProvide /llms.txt as forward-looking hygiene.
Sourcellmstxt.org spec v2; Ahrefs llms.txt crawl study (2026).
Freshness 20
Whether the page signals it is current — dates, sitemap, canonical URL, and content depth.
Content Freshness Date
5 ptsLabel content update dates so AI knows it is current.
SourceDocumented AI-engine preference for recently updated content.
XML Sitemap
5 ptsProvide an XML sitemap to help crawlers discover pages.
SourceSitemap protocol.
Canonical URL
4 ptsSet a canonical URL to avoid duplicate content issues.
SourceCanonical link annotation standard.
Content Depth
6 ptsEnsure enough content depth — thin pages are rarely cited.
SourceCorrelation measured across the audit corpus.
Layer-3 Probe Method
Layer 3 (AI Search Visibility Probe) uses a different method: real buyer-style questions are generated by an LLM and are verboten from containing the brand name, then asked to multiple real models concurrently. Brand identification prefers deterministic ordered-list parsing over model self-reporting; Perplexity citations are trusted natively while other engines' citations must cross-match URLs in the answer text. Scores combine mention rate (60%), ranking quality (25%) and sentiment (15%).
Changelog
- · Added: AI Crawler Reachability (edge/blocking detection), Content Citability (quotes & statistics), Bing Index Coverage.
- · Expanded the AI bot list from 8 to 14 purpose-tagged user agents.
- · De-emphasized llms.txt honestly (weight 4 → 2) after crawl studies showed 97% of files receive zero requests.
- · Weights redistributed within the same four categories; total remains 100.
- · Initial 18-check rule set across four categories (crawlability / content / authority / freshness), 0-100 scoring with pass/warn/fail semantics (full / half / zero points).