Domain Authority does not predict whether AI systems will cite your site. A first-party baseline audit using the AI Visibility Readiness (AVR) Framework showed citability.dev achieving a 15% AI citation rate while ahrefs.com measured 5% on the same 20-query protocol.
Why this matters #
citability.dev (a new, low-authority domain) achieved a 15% AI citation rate in a 20-query ChatGPT baseline test run April 2026. ahrefs.com achieved 5% on the same protocol tested March 2026. The AI Visibility Readiness (AVR) Framework produced both measurements with 95% confidence intervals; both results carry LOW confidence labels due to small n. The finding supports the hypothesis that AI infrastructure choices matter more than domain authority for AI discoverability.
citability.dev achieved a 15% AI citation rate on a 20-query ChatGPT baseline while ahrefs.com, one of the most authoritative SEO domains in the world, achieved 5% on the same protocol run one week earlier. The finding is direct: AI infrastructure choices matter more than domain authority for AI discoverability, and the gap is measurable with the AVR Framework.
The standard playbook for search authority says high Domain Authority equals search visibility. The new SEO reality — where ChatGPT, Perplexity, and Google AI Overviews now answer queries that used to send traffic to article pages — does not follow that playbook.
In April 2026, I ran a citation-rate baseline for citability.dev using the AI Visibility Readiness (AVR) Framework (see chudi.dev/framework), a first-party audit methodology I built to measure whether a website is discoverable, recommendable, and citable by AI systems. The result: citability.dev was cited in 3 of 20 ChatGPT queries — a 15.0% citation rate.
For comparison, I had run the same 20-query citation protocol on ahrefs.com one week earlier, on March 31, 2026. Ahrefs — one of the most authoritative SEO tool domains in the world — was cited in 1 of 20 queries: a 5.0% citation rate. This document is the methodological record for both measurements.
The measurements #
citability.dev citation baseline (April 7, 2026)
- Queries run: 20
- Platform: ChatGPT (OpenAI), web search enabled
- Cited: 3
- Citation rate: 15.0%
- 95% confidence interval: [5.2, 36.0] percentage points
- Confidence label: LOW
- Verdict: PARTIALLY_CITED
source: `avr-baseline-2026-04-07/citations_citability.dev_20260407_074025_summary.json`
, field citation_rate_pct
, test_date 2026-04-07T07:40:25Z
ahrefs.com citation baseline (March 31, 2026)
- Queries run: 20
- Platform: ChatGPT (OpenAI), web search enabled
- Cited: 1
- Citation rate: 5.0%
- 95% confidence interval: [0.9, 23.6] percentage points
- Confidence label: LOW
- Verdict: PARTIALLY_CITED
source: sample-audits/citations_ahrefs.com_20260331_160501_summary.json
, field citation_rate_pct
, test_date 2026-03-31T16:05:01Z
What “LOW confidence” means
Both results carry a LOW confidence label. With n=20 queries, the 95% Wilson confidence intervals are wide — roughly plus or minus 15 to 20 percentage points. The citability.dev interval [5.2, 36.0] and the ahrefs.com interval [0.9, 23.6] overlap at the edges. What the data supports is directional: citability.dev is being cited where ahrefs.com is not, on queries in the same broad topic space. What it does not support is a precise ratio claim. Tight confidence intervals require larger n; the AVR protocol documentation recommends n=200 for a meaningful operational baseline.
What drove the difference #
The AVR Framework distinguishes between VERIFIABLE checks (infrastructure you can prove) and BEST-EFFORT checks (measured behavior that varies by session). The ahrefs.com infrastructure audit, also run in March 2026, found several gaps that correlate with reduced AI discoverability: no /llms.txt
, no schema markup on 5 of 5 checked pages (0% coverage), and partial semantic HTML. These are all VERIFIABLE findings — they do not require querying an AI to observe.
citability.dev, by contrast, was built from the start with AI-visible content structure in mind: structured data on key pages, clear heading hierarchy, llms.txt
describing the service, and explicit AI crawler allowlisting in robots.txt
.
The AVR Framework also measured citability.dev’s visibility rate separately from its citation rate. In a 25-query visibility test on April 7, 2026:
- Visible in 11 of 25 queries: 44.0% visibility rate
- Of which, brand recognition (5/5 queries, 100%), concept attribution (2/10, 20%), and recommendation (4/10, 40%)
source: `avr-baseline-2026-04-07/visibility_citability.dev_20260407_074010_summary.json`
, fields visibility_rate_pct
, by_category
What the framework measures #
The AVR Framework v1.1.0 (repository: github.com/ChudiNnorukam/ai-visibility-readiness), which I also ran against chudi.dev in a full 13-section self-audit with Cloudflare Radar baselines, runs up to 15 automated checks across two sections:
Section 1 (VERIFIABLE): SEO Foundation Technical crawlability, schema markup coverage, Core Web Vitals, content indexability, content-to-HTML ratio, and semantic HTML structure. These checks produce deterministic pass/fail results from public data.
Section 2 (VERIFIABLE): AI Infrastructure AI crawler access permissions (GPTBot, ClaudeBot, PerplexityBot, Google-Extended), /llms.txt
presence, structured data depth across 5 sampled pages, content structure for AI parsing, and semantic HTML depth.
Section 3 (BEST-EFFORT): Citation Monitoring Live citation tests across ChatGPT and optionally Perplexity, reported with n, citation rate, and 95% Wilson confidence intervals. Results carry a mandatory confidence label (LOW / MODERATE / HIGH based on round count) and the disclaimer that AI citation behavior varies by session, platform, and location. These are point-in-time observations, not audit verdicts.
The three-band verdict (AI-READY / FOUNDATION-READY / NOT-READY) is driven by Section 1 and 2 findings only. It is not a composite score; it is a categorical readiness classification.
Replicating these measurements #
Both citation baselines were produced by running scripts/citation_auto.py
in the AVR repository against the target domain with a set of 20 domain-appropriate queries. The script queries ChatGPT via the OpenAI API with web search enabled, checks whether the target URL appears in each response’s citations, and outputs the summary JSON with Wilson CI.
To run your own baseline:
- Clone
github.com/ChudiNnorukam/ai-visibility-readiness - Install dependencies (
pip install -r requirements.txt
) - Set
OPENAI_API_KEY
in your environment - Run
`python scripts/citation_auto.py <your-domain> --brand "Your Brand" --n 20`
For a meaningful operational baseline, n=200 is recommended. n=20 is the pilot scale: it validates the measurement pipeline and produces directional signal, not a precise rate.
Implications #
AI citation does not inherit from search authority. A domain can have strong traditional SEO signals — high DA, excellent Core Web Vitals, strong backlink profile — and still register near-zero citation rates in AI systems if the content structure and crawler infrastructure do not meet what AI systems need.
The inverse is also measurable: a new domain with deliberate AI infrastructure choices can achieve citation rates that established, high-authority domains do not.
This is the core insight citability.dev operationalizes: audits measure what you actually get from AI systems, not what traditional SEO metrics predict you should get. The chudi.dev self-audit and the five-page remediation arc document exactly what changed to close the infrastructure-to-citation gap.
All measurements in this post are first-party, run with the AVR Framework on the dates stated. Both results carry LOW confidence labels; n=20 is insufficient for high-confidence rate estimation. For questions about the methodology, see chudi.dev/framework or the repository FRAMEWORK.md.
· Sources & further reading
Sources & Further Reading #
Further reading
Cloudflare Will Block AI Crawlers by Default on September 15: What Site Owners Need to Do Now /blog/cloudflare-block-ai-crawlers-september-15Cloudflare blocks AI crawlers by default on September 15, 2026. Should you block, allow, or charge via Pay Per Crawl? Decision matrix + verification steps.I Audited My Own Site With AVR v1.1.0. Here Is What I Found. /blog/avr-v1-1-0-case-study-auditThe first comprehensive 8-section AVR v1.1.0 audit of chudi.dev produced AGENT-READY 3/3 on §2.7 but only 40/100 on Fact-Block Density. Here is the full audit, plus the 2026-06-03 update where remediation lifted the root and framework pages to a verified 100/100 EXTRACTABLE.AVR v1.1.0 Full 13-Section Audit: Cloudflare Radar Data Meets AI Visibility /blog/avr-v1-1-0-cloudflare-radar-full-auditThe first complete AVR v1.1.0 audit with all 13 sections, including 5 new Cloudflare Radar-derived checks. chudi.dev passes every Cloudflare-backed check. Here are the numbers.How I Lifted Five chudi.dev Pages to EXTRACTABLE on AVR v1.1.0. /blog/avr-v1-1-0-five-page-remediation-arcAll 5 audited chudi.dev URLs now score EXTRACTABLE on AVR v1.1.0 Fact-Block Density. Two HTML traps (dt/dd Q/A pairs and icons before heading text) cost a follow-up commit each. The CI workflow now hard-fails any regression.How to Get Cited by ChatGPT: My 0/5 GEO Audit /blog/how-to-get-cited-by-chatgpt-geo-guideI ran my own site through 5 buyer-intent queries on Perplexity and got cited zero times. Here is what GEO actually is and the pattern the winners share.
What do you think? #
I post about this stuff on LinkedIn every day and the conversations there are great. If this post sparked a thought, I'd love to hear it.