AI Visibility Readiness, a transparent, tiered audit methodology for measuring whether your site is readable, recommendable, and citeable by traditional and AI-powered search systems. Every check ships with an evidence label.
AVR does not invent composite scores for things it cannot measure. Every check in the framework is tagged with how much you can trust the result. Verifiable checks are reproducible against a free API or CLI tool. Best-effort checks measure a point-in-time sample whose confidence is labelled explicitly per round.
Objective, reproducible, backed by a free API or CLI tool. Anyone can run the check and get the same result.
Measurable in a point-in-time sample but varies by session, query phrasing, and platform state. Confidence labelled explicitly.
AVR v1.1.0 runs eight audit sections against a target URL. Sections 1 through 5 cover the traditional SEO foundation, AI infrastructure files, live citation polling, AI visibility testing, and a calibration receipt. Sections 6 through 8 are the v1.1.0 additions for Agent Readiness, Fact-Block Density, and Citation Decay Rate.
The baseline AI search systems depend on. If organic search cannot find you, AI search never sees you.
The AI-specific files that declare your intent to crawlers. llms.txt reclassified to optional/documentation-only per Google I/O 2026; Lighthouse 13.3 flags it as ERROR by default.
Does AI link to your URL when asked about your topics? 20 queries across ChatGPT, Perplexity, Claude, Gemini. Single rounds are LOW confidence.
Does AI know you exist and recommend you? Brand recognition + concept attribution + recommendation signals. With AI Mode at 1B+ monthly users and AI Overviews dropping CTR 58%, this is now the primary arena.
Pre-flight smoke test proving the audit tool was working correctly when it produced the numbers. The forensic trail unit that makes the deliverable defensible.
Does your site expose tools to AI agents? WebMCP manifest + A2A AgentCard + wire-protocol probe. The browser-level standard shipping in Chrome 146 Canary; W3C draft co-authored by Google + Microsoft.
Is your content shaped for citation? Five weighted checks score 0-100. Patel's first-sentence-of-every-H2 rule + AI SEO Engineering Standards 40-60w direct-answer band. The proximate driver of citation selection.
How long do you HOLD a citation? The moat metric no other AI visibility tool tracks. Consumes Bing AI Performance CSV; computes half-life, decay slope, displacement events, retention rate.
The framework collapses the eight section verdicts into one of three tiers. No fake composite score, just the honest read on where the site lives today. AI-READY requires foundation plus AI infrastructure plus measurable citations. Foundation-ready means infrastructure mostly in place with gaps. Not-ready means foundation itself fails and AI work is premature.
Sections 1+2+6+7 PASS + measurable citations (>0% per round) within last 30 days + EXTRACTABLE Fact-Block + GROWING/STABLE Citation Decay
Sections 1+2 PASS + Sections 6+7 PARTIAL. Infrastructure in place; gaps in agent-readiness or content extractability.
Section 1 FAIL OR Section 2 FAIL. AI infrastructure is upstream; foundation must come first.
chudi.dev runs its own AVR v1.1.0 audit in public, because transparency is the differentiator. The site lands AI-READY overall, with AGENT-READY 3 of 3 on Section 6 and a verified 100 of 100 EXTRACTABLE on Section 7 Fact-Block Density. The numbers below are this page's own most recent AVR run.
After spending 106 hours and $10K on AEO infrastructure that produced zero AI citations, the lesson was clear. The infrastructure was correct. The SEO foundation underneath it did not exist. AVR exists so the next site builder runs the checks in the right order and skips the same mistake.
The framework enforces the right order: foundation first, then AI readiness, then live AI testing. Running visibility tests before the foundation is in place is premature.
Material changes to AVR are versioned, editorial fixes and link updates are not. The v1.1.0 release adds three new audit sections (Agent Readiness, Fact-Block Density, Citation Decay Rate) and reclassifies llms.txt from required to optional per Google I/O 2026 guidance. The earlier v1.0 to v1.1.0 jump on 2026-04-18 added Section 4 separately.
Sections 6 (Agent Readiness, WebMCP + A2A AgentCard), 7 (Fact-Block Density, content extractability scoring), and 8 (Citation Decay Rate, the moat metric) added. llms.txt reclassified to optional/documentation-only per Google I/O 2026.
Section 4 (AI Visibility) added. Concept attribution and recommendation rate now scored separately from raw citation count. Reflects the gap between "AI knows the brand" and "AI cites the URL."
Initial release. Three sections (SEO Foundation, AI Infrastructure, Citation Monitoring) plus three-band verdict.
The five questions every prospect asks about AVR v1.1.0 before running it on their site.
scripts/requirements.txt, and run python3 scripts/run_audit.py YOUR_URL --skip-lighthouse --full-v11. Free audit covers Sections 1, 2, 6, 7, 8 at zero API cost. Add --live-test with brand and topics for the paid Sections 3, 4, 5 (about two dollars in API spend per audit)./.well-known/webmcp exists and validates as JSON with at least one tool declared. Second, /.well-known/agent.json exists with required AgentCard fields. Third, each endpoint the manifests declare resolves to a non-404 response. Three of three means all pass. chudi.dev and citability.dev both reach this verdict.The tool powering citability.dev runs AVR v1.1.0 against any URL. The free scan covers eight v1.1.0 sections at zero API cost. The full audit adds live AI engine polling across OpenAI, Perplexity, Claude, Gemini, plus a calibration receipt for the audit, costing about two dollars per run.