A New Domain Outscored Ahrefs on AI Citation Rate. Here Is How We Measured It.

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) 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:

1. Clone github.com/ChudiNnorukam/ai-visibility-readiness
2. Install dependencies (pip install -r requirements.txt)
3. Set OPENAI_API_KEY in your environment
4. 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.

---

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.