ai-citation-auditor

AI Citation Auditor

Will AI cite this page? A free, dependency-free Claude skill that audits any page, predicts whether AI answer engines will cite it, and returns a prioritized fix list.

Ranking in the ten blue links and getting cited in an AI Overview are now two different games. This skill scores the second one. It runs entirely on tools any Claude has out of the box (web fetch, web search, or content you paste in). No paid SEO stack, no crawler, no API keys.

Why this exists

Search is moving from ten blue links to synthesized AI answers across Google AI Overviews and AI Mode, ChatGPT, Perplexity, Gemini, and Claude. Most teams still measure rankings and nothing else. A page can sit at position one and get skipped by the AI answer sitting right above it. This skill tells you why, and what to fix.

The mental model: citation is a vote

An AI answer engine does not rank sources the way classic Search does. For every answer it runs a silent vote between competing sources of truth:

1. Latent - what the model already learned in training. Wins by default when nothing contradicts it.
2. Retrieval - what the model just fetched for this query. Overrides latent for anything current.
3. Source-quality filter - a hidden re-ranker that discounts aggregators, thin listicles, forums, and pages that read like manipulative SEO, before the vote even happens.

There is a hard prerequisite underneath all three: the model has to be able to retrieve you at all. Retrieval rank is still the strongest predictor of AI citation. Fix retrievability first, then compete on the layers.

What it does

AUDIT (default): give it a live URL or pasted content plus a target query. It returns a 0 to 100 score, a per-layer breakdown, a verdict, and a ranked fix list.

REWRITE: give it one section plus the query it should answer. It returns a citation-ready version (answer-first, self-contained, consensus-aligned).

Scoring rubric (6 layers)

PASS at 80 or above. MARGINAL 60 to 79. FAIL under 60. Any critical layer under 40 forces FAIL.

Evidence tiers (no black-box scores)

Every signal is labeled so you know what is solid and what is directional:

• Confirmed - stated in Google's own published Search and AI guidance, or standard practice.
• Leak-inferred - drawn from the 2024 Google Content Warehouse API documentation leak. Directional.
• Model-observed - drawn from how answer engines behave in the field and from published model guidance. Directional and subject to silent change.

Install and run

Option A (fastest): paste the contents of SKILL.md into a Claude Project's custom instructions.

Option B: drop the ai-citation-auditor/ folder into your Claude skills directory.

Then ask:

Run an AI citation audit on https://example.com/page for the query "your target query"

Example output

A live audit of a top-ranking "best password manager" page. Genuinely well made (named cybersecurity author, hands-on testing, current date), and it still lands MARGINAL because its headline pick fights the web's consensus:

AI CITATION AUDIT
Page:          passwordmanager.com/best-password-managers/
Target query:  "best password manager"

VERDICT:  MARGINAL (68 / 100)      bands: PASS >=80 | MARGINAL 60-79 | FAIL <60
Retrieval prerequisite:  surfaces in top results? YES

LAYER                           SCORE  WT   TIER              DIAGNOSIS
1  Answer-first extractability   78   22%  Confirmed/Model   H1 answers in sentence one; clean Q&A blocks
2  Evidence and corroboration    68   20%  Confirmed         First-hand testing, but scores + audit claims uncited
3  Consensus alignment           55   16%  Model-observed    Crowns RoboForm #1; consensus leans 1Password/Bitwarden
4  Source-quality fingerprint    66   18%  Leak-inf/Conf     Real author + disclosure; heavy affiliate posture
5  SEO-slop detection            62   12%  Leak-inferred     Professional, but templated blocks + long generic FAQ
6  Machine structure             72   12%  Confirmed         Strong H-tree, tables, FAQ block; JSON-LD not verifiable

What this skill does NOT do

• It does not write full articles. It audits and rewrites single sections.
• It does not build links or off-site authority.
• It does not track rankings. The prerequisite check is a directional read from a live search.
• It does not guarantee a citation. Answer-engine logic is undocumented and shifts silently. Treat scores as decision support.

License

Free to use, modify, and share. Attribution appreciated.

Author

Niro, Founder of GODRANK (Madrank Digital Ltd). X: @PGR_mx

Built from a larger internal retrieval-layer auditing system, stripped down and open-sourced so anyone can run it with no setup.