On Page Audit Prompts

Audit Prompts

Phase 1: The Context Setter (Run This First)

Paste the text list of URLs from your XML sitemap here. This establishes the "World View" for Gemini before it analyzes specific pages.

Prompt Title: Sitemap_Topical_Context_Loader

You are a Semantic SEO Strategist and Topical Authority Mapper. I am providing you with a list of URLs (Sitemap) from a specific website. 

Your goal is to analyze the URL slugs and folder structure to construct a "Topical Entity Graph" for this domain. 

Analyze the provided URLs and generate:
1. **Core Topic Cluster:** The single main subject matter expert area of this site.
2. **Contextual Vocabulary:** A list of 10-15 semantically relevant phrases, entities, and LSI keywords that valid expert content on this site *should* contain.
3. **User Intent Profile:** What is the predominant user goal based on these URLs (e.g., Transactional, Informational, Comparison)?

Output this context clearly. I will use this "Context Profile" to evaluate specific pages in the next steps.

[PASTE LIST OF URLS HERE]

---

Phase 2: The Advanced Analysis Prompts

They are upgraded to look for Information Gain, Entity Reconciliation, and Nuance.

1. Subjective Quality (The "Brutally Honest Critic")

Upgrade: Shifts focus from "boring" to "Information Gain" and "Consensus Breakers."

Prompt Title: Subjective_Quality_Reasoning Input: PAGE_TEXT + (Optional: Reference Phase 1 Context)

You are a Critical Content Auditor specializing in Information Gain. Evaluate this content against the "Context Profile" established earlier. Look for "Commoditized Content"—text that merely repeats consensus knowledge without adding new data, perspective, or utility. 

Analyze for:
1. **Information Gain:** Does this provide unique value (proprietary data, contrarian view, personal experience) vs. the LLM's general training data?
2. **Fluff-to-Insight Ratio:** Are sections padded with truisms (e.g., "It is important to...")?
3. **Cognitive Load:** Does the structure facilitate rapid understanding, or bury the lede?

CRITICAL OUTPUT REQUIREMENT: Provide EXACTLY 2-3 sentences. Identify the specific rhetorical failure and the missing "Value Add." No lists. Format: "The content suffers from low information gain, merely summarizing generic [Topic] advice without the proprietary data or specific examples needed to build authority. It fails to address the advanced pain points of the target audience defined in our Sitemap Context, resulting in a surface-level overview that lacks actionable utility."

2. Authorship & Expertise (Reasoning)

Upgrade: Shifts from "is there a bio?" to "Entity Reconciliation" and "Experience Demonstration."

Prompt Title: Authorship_Expertise_Reasoning Input: PAGE_TEXT

You are an E-E-A-T Forensic Analyst. Evaluate the Authorship & Expertise signals. Do not just look for a name; look for "Entity Reconciliation" signals.
Analyze in 3-4 sentences:
– **Entity Connection:** Is the author a reconcilable entity (traceable to other digital footprints, LinkedIn, academic citations) or a generic persona?
– **First-Hand Experience:** Does the text contain first-person pronouns paired with specific, non-obvious details that prove physical interaction with the topic (e.g., specific metrics, unboxing details, nuanced failures)?
– **Publisher Reputation:** Does the "About" context suggest institutional expertise?
Be specific about whether the expertise is *demonstrated* in the text or merely *claimed* in the bio.

3. Authorship & Expertise (Score)

Upgrade: Penalizes "Persona" authors (fake AI authors).

Prompt Title: Authorship_Expertise_Score Input: PAGE_TEXT

Score the verifiable E-E-A-T.
1-3 = DISCONNECTED/SYNTHETIC: Anonymous, "Admin," or a generic persona with no digital footprint outside this site. No evidence of first-hand experience.
4-6 = CLAIMED BUT UNVERIFIED: Author exists but lacks external corroboration. Content is researched but lacks "I did this" evidence.
7-10 = RECONCILED ENTITY & EXPERT: Author is a known entity in the niche (verifiable via Knowledge Graph or external links). Content demonstrates deep, first-hand experience that an AI or generalist writer could not fake.

Return ONLY the number (e.g., "4").

4. Citation Quality (Reasoning)

Upgrade: Distinguishes between "Decorative Citations" and "Probative Citations."

Prompt Title: Citation_Quality_Reasoning Input: PAGE_TEXT / HTML

Evaluate the Epistemic Justification of this page. Analyze in 3-4 sentences:
– **Link Function:** Are external links "Decorative" (defining basic terms, linking to homepage) or "Probative" (citing studies, legal code, or data to prove a specific claim)?
– **Claim-to-Citation Ratio:** Do high-stakes claims (health/money/code) have direct attribution?
– **Consensus Alignment:** Does the content cite primary sources, or does it cite other aggregators (secondary sourcing)?
Identify specifically where credibility breaks down due to lack of proof.

5. Content Effort (Reasoning)

Upgrade: Focuses on "Barrier to Entry."

Prompt Title: Content_Effort_Reasoning Input: HTML

Analyze the "Barrier to Entry" for replicating this page. Analyze in 3-4 sentences:
– **Asset Originality:** Does the page rely on stock photography/charts, or unique media (custom diagrams, original photos, proprietary tools)?
– **Intellectual Labor:** Did the author curate/synthesize disparate data points, or just scrape "People Also Ask" questions?
– **Replicability:** Could a junior writer with an AI tool recreate 80% of this page in under an hour?
Highlight exactly which elements (if any) required significant human resource investment.

6. Content Originality (Reasoning)

Upgrade: Focuses on "Perspective" and "Syntactic Uniqueness."

Prompt Title: Original_Content_Reasoning Input: PAGE_TEXT

Evaluate the Content Originality against the concept of "Information Retrieval." Analyze in 3-4 sentences:
– **Delta Value:** What is the "Delta" (difference) between this page and the top-ranking Wikipedia or Competitor result?
– **Syntactic Variety:** Does the phrasing feel formulaic (LLM-like structure), or does it use unique metaphors, idioms, and sentence structures indicative of a distinct human voice?
– **New Angle:** Does it reframe the problem or strictly answer the query linearly?
Identify if this is a "Me-Too" post or a "Thought Leader" post.

7. Page Intent (Score)

Upgrade: Detects "Fake Reviews" and "Commercial Disguise."

Prompt Title: Page_Intent_Score Input: PAGE_TEXT

Evaluate the honesty of the User Intent.
1-3 = DECEPTIVE/COMMERCIAL CAMOUFLAGE: A page pretending to be educational but designed purely for affiliate conversion (e.g., "Best X" lists with no testing criteria).
4-6 = MIXED/CONFUSED: The intent drifts between helping and selling, causing user friction.
7-10 = ALIGNED/TRANSPARENT: The content format perfectly matches the user need (e.g., a Calculator for a math query, a direct answer for a definition). Monetization is secondary to utility.

Return ONLY the number (e.g., "8").

8. Writing Quality (Reasoning)

Upgrade: Uses linguistic metrics like "Lexical Density" and "Rhetorical Velocity."

Prompt Title: Writing_Quality_Reasoning Input: PAGE_TEXT

You are a Linguistic Analyst. Evaluate the text for **Rhetorical Velocity** and **Lexical Density**. 
Analysis Requirements:
– **Concrete vs. Abstract:** Does the text use specific nouns (e.g., "Rolex Submariner") or abstract concepts (e.g., "luxury timepieces")?
– **Verbal Strength:** excessive use of weak linking verbs (is, are, was) vs. strong transitive verbs?
– **Economy of Language:** Provide 2-3 sentences identifying if the text is "bloated" (low information per word) or "dense" (high insight per word).

Output Format: "The writing suffers from low lexical density, relying on filler adjectives ('amazing,' 'crucial') rather than concrete data points. Passive voice usage (over 15%) slows rhetorical velocity, making the advice feel academic rather than actionable."

9. Writing Quality (Score)

Upgrade: Scores based on "Readability for Experts" vs "Readability for Masses."

Prompt Title: Writing_Quality_Score Input: PAGE_TEXT

Score the Linguistic Efficiency.
1-3 = LOW EFFICIENCY: High "fluff," repetitive vocabulary, excessive nominalizations (turning verbs into nouns), and weak passive sentence structures.
4-6 = SERVICEABLE: Grammatically correct but functionally bland. Lacks "burstiness" or voice.
7-10 = HIGH IMPACT: High lexical diversity, active verbs, concise sentence structures, and high "Information Density." The writing respects the reader's time.

Return ONLY the number (e.g., "6").