I Audited My Own Site With AVR v1.1.0. Here Is What I Found.

The AI Visibility Readiness Framework moved to v1.1.0 last week. The new spec adds three audit sections nobody else tracks: WebMCP Agent Readiness (§2.7), Fact-Block Density (§3.7), and Citation Decay Rate (§7). I ran the full audit against chudi.dev on 2026-05-23 at 07:00 UTC. Here is the report.

The headline numbers

chudi.dev passed SEO Foundation (Section 1), AI Infrastructure (Section 2), and Agent Readiness (Section 6). It landed AGENT-READY 3/3 because the WebMCP manifest at /.well-known/webmcp resolves, the A2A AgentCard at /.well-known/agent.json resolves, and the five API endpoints those manifests declare (/api/webmcp/list-posts, /api/webmcp/read-post, /api/webmcp/avr-self-audit, /api/webmcp/author-profile, /api/agent/dispatch) all return real handlers with valid CORS preflights. Citation Decay Rate came back GROWING with retention 1.65 across 89 days of Bing AI Performance data and the post-tune half-life algorithm correctly returned no half-life crossing.

Fact-Block Density failed. chudi.dev root scored 40 out of 100 (NOT-EXTRACTABLE). The framework page scored 51 out of 100 (PARTIALLY-EXTRACTABLE).

Why Fact-Block Density failed even though the rest passed

The infrastructure checks ask whether AI engines CAN reach the content. The Fact-Block Density checks ask whether the content they reach is shaped for citation. Those are different questions.

On chudi.dev/framework the audit found that 100% of H2 sections open with sentences that stand alone as answers, but 0% of those H2 paragraphs land in the 40-60 word direct-answer band. The opening sentence is right; the supporting paragraph is the wrong length. Only 8.3% of H2 and H3 headings are phrased as questions a user would actually type. There is no FAQ section at the bottom of the page.

Three failures, three remediation targets. The first-sentence rule already passes. The path from 51 to 75+ is not a content rewrite, it is a paragraph-length tune and a heading rewrite and an FAQ addition.

The remediation plan

Read the full plan at ai-visibility-readiness/AVR-V1-1-0-FACT-BLOCK-REMEDIATION-PLAN.md in the repo. The headline targets:

Page	Current score	Top failing check	Fastest remediation
chudi.dev root	40/100	direct-answer band + question headings + FAQ	Add FAQ section; rewrite H2 openings to 40-60 words; convert top 3 H2 headings to questions
chudi.dev/framework	51/100	direct-answer band + question headings + FAQ	Same three remediations applied to the methodology bento sections
/blog (index)	not yet audited	TBD	Run section_fact_block_density.py on this URL
/work (index)	not yet audited	TBD	Same
/about	not yet audited	TBD	Same

Each rewrite is a 5 to 15 minute edit per H2 section. The framework page has 4 main H2 sections, so the lift is 20 to 60 minutes total. The score should move from 51 to 75+ once the band check passes and the FAQ lands.

Why this is a defensible deliverable

Most AI visibility tools return a score and nothing else. The AVR v1.1.0 audit returns five things you can act on: the verdict for each of 8 sections, the per-check pass/fail with compliance percentages, the specific recommendation each failing check generates, the canonical evidence the recommendation cites, and the reproducible command you can re-run to verify the fix landed.

For Fact-Block Density that means I know exactly which 92% of headings on chudi.dev/framework are not in question format, exactly which 100% of H2 paragraphs miss the 40-60 word band, and I have a runnable command (python3 run_audit.py chudi.dev/framework --skip-lighthouse --full-v11) that will confirm when remediation lands the score above 75.

For Citation Decay Rate that means I know the algorithm correctly suppressed the previous false-positive (21-day half-life reading on overall-growing data) because the 12-case pytest regression suite validates the suppression. The retention rate of 1.65 across 89 days is computed from real Bing CSV data, not estimated.

The case-study hypothesis ratifies

chudi.dev is supposed to DEMONSTRATE AVR. citability.dev is supposed to IMPLEMENT AVR. Until this week both claims were aspirational. As of 2026-05-23 both sites land AGENT-READY 3/3 on production with meaningful endpoint resolution. Both produce real JSON responses on the declared WebMCP tools. Both pair with the same audit script that produces the same shape of report.

The piece left is Fact-Block Density remediation on chudi.dev pages. That is a content-editing arc, not an infrastructure arc. It is the next thing I will ship.

Update, 2026-06-03: the remediation landed

The remediation shipped, and it overshot the prediction. The numbers in the audit above are the documented starting point. These are the after.

I re-ran the same section_fact_block_density.py against the live pages, through the automated post-deploy audit that now runs on every chudi.dev deploy. The results:

Page	Before (2026-05-23)	After (2026-06-03)
chudi.dev root	40/100 NOT-EXTRACTABLE	100/100 EXTRACTABLE (5/5)
chudi.dev/framework	51/100 PARTIALLY-EXTRACTABLE	100/100 EXTRACTABLE (5/5)
chudi.dev/blog	not yet audited	100/100 EXTRACTABLE (5/5)
chudi.dev/about	not yet audited	85/100 EXTRACTABLE (4/5)
chudi.dev/topics	not yet audited	80/100 EXTRACTABLE (4/5)

The fix was exactly the three rules the plan named, nothing more: rewrite each H2 opening paragraph into the 40-60 word direct-answer band, phrase the headings as questions, and add an FAQ section. No re-architecture. The predicted path was 51 to 75+. The real result was a perfect 100/100. The two pages still short of it, /about at 85 and /topics at 80, already clear the EXTRACTABLE bar and have one check each left to tune.

One honest footnote, because it is the real lesson. When I went to verify the framework page, it was still publicly displaying its old 40/100 score in the results card, flagged “remediation queued,” even though the content had already been remediated to pass. The score is not a trophy you win once. It drifts the moment you stop re-measuring. That stale number is corrected now too, which is the only reason this update exists.

Run the audit yourself

git clone https://github.com/ChudiNnorukam/ai-visibility-readiness.git
cd ai-visibility-readiness
pip install -r scripts/requirements.txt
python3 scripts/run_audit.py YOUR_URL --skip-lighthouse --full-v11

The free audit gives you Sections 1+2+6+7+8 at $0 cost. Add --live-test --brand "..." --owner "..." --topics "..." for Sections 3+4+5 (live AI engine polling, ~$2 per audit).

The report drops into sample-audits/audit_<domain>_<timestamp>.md. The per-section JSON files drop alongside for programmatic analysis. The pytest suite at tests/test_section_citation_decay.py runs in 0.04 seconds.

If your site comes back NOT-EXTRACTABLE on Fact-Block Density and you do not know where to start, the recommendations in the audit JSON tell you which H2 sections fail which checks and in what order to fix them. That is the deliverable.