This protocol is a manually/AI-executable GEO detection checklist that does not rely on any external tools. It is applicable for detecting the "AI citation friendliness" of a single article, outputting a structured detection report, and providing actionable optimization recommendations.
| Principle | Description |
|---|---|
| Locally Executable | |
| All detection items can be completed without calling external APIs, manually verifiable | |
| Modular Configuration | |
| 11 detection modules can be dynamically toggled on/off as needed | |
| Quantifiable Scoring | |
| Each module outputs a score of 0-100, generating a comprehensive GEO score | |
| Actionable Optimization | |
| Each low-scoring item is accompanied by specific optimization guidance | |
| Aligned with Existing Frameworks | |
| Integrates design thinking from the 13-module writing template, engagement depth lens, and metaphor-quote formula |
This protocol is designed based on the following core GEO principles:
| Principle | Data Source |
|---|---|
| Content with high trustworthiness is 3.7 times more likely to be cited by AI | |
| β | |
| Adding FAQ Schema increases AI citation rate by approximately 47% | |
| β | |
| The same information cross-verified across multiple trusted sources reduces single-point pollution risk | β |
| Core GEO metrics: Visibility, Source Citation Rate, Information Correction Rate, Expression Richness, Information Balance, Priority Recommendation Rate | |
| β | |
| Four core metrics: Semantic Embedding Depth, Authority Accumulation, Generative Contribution, Transformation Drive | |
| β | |
| E-E-A-T (Experience, Expertise, Authoritativeness, Trustworthiness) is the gold standard for content quality in the AI era | β |
Detection Goal: Confirm whether AI crawlers can normally access the article content
| Item ID | Check Item | Method | Weight |
|---|---|---|---|
| T1 | Is the article URL publicly accessible? | Open the article link in an incognito/private browser window | High |
| T2 | Is the article content readable as plain text (not images/PDF)? | Check if the page is HTML text | High |
| T3 | Does the article have a clear <title> tag? |
||
| Check the browser tab title | Medium | ||
| T4 | Does the article have a <meta description> ? |
||
| View page source or browser description | Medium | ||
| T5 | Does robots.txt allow major AI crawlers (GPTBot/ClaudeBot/Bytespider/etc.)? | Check the /robots.txt file |
|
| High |
Scoring Criteria: 20 points per item, maximum 100 points.
Detection Goal: Confirm whether the article uses structured markup understandable by AI
| Item ID | Check Item | Method | Weight |
|---|---|---|---|
| S1 | Is JSON-LD structured data used? | Look for <script type="application/ld+json"> in page source |
|
| High | |||
| S2 | Does it contain Article Schema markup? |
||
Look for @type: "Article" in JSON-LD |
|||
| High | |||
| S3 | Does it contain FAQPage Schema markup (if applicable)? |
||
Look for @type: "FAQPage" in JSON-LD |
|||
| Medium | |||
| S4 | Does it contain Author Schema markup? |
||
| Look for author information in JSON-LD | Medium | ||
| S5 | Does the Schema contain entity relationship markup like sameAs ? |
||
Look for sameAs , knowsAbout , etc. in JSON-LD |
|||
| Low |
Scoring Criteria: 20 points per item, maximum 100 points.
Theoretical Basis: Adding FAQ Schema can increase AI citation rate by approximately 47%. Schema markup helps AI understand the type and relationships of content.
Detection Goal: Confirm whether the article structure facilitates AI information extraction
| Item ID | Check Item | Method | Weight |
|---|---|---|---|
| C1 | Is there a clear H1 heading? | Check if the main title uses # or H1 |
|
| High | |||
| C2 | Are there hierarchical H2/H3 subheadings? | Check if the article has 2-3 levels of heading structure | High |
| C3 | Is there an abstract/introduction (within 200 words)? | Check if there is a summary paragraph at the beginning | High |
| C4 | Are lists/tables used to organize information? | Check for <ul> , <ol> , or <table> in the article |
|
| Medium | |||
| C5 | Does each H2 section have at least 300 words of supporting content? | Estimate word count per section | Medium |
| C6 | Is there a 40-60 word section summary? | Check if each section ends with a summary paragraph | Low |
Scoring Criteria: C1-C3 = 20 points each, C4-C6 = 13.3 points each, maximum 100 points.
Detection Goal: Confirm whether the article's core concepts are clearly recognized by AI
| Item ID | Check Item | Method | Weight |
|---|---|---|---|
| E1 | Are core concepts clearly defined in the text? | Check if key terms are explained at first appearance | High |
| E2 | Are core concepts used consistently throughout? | Check if the same concept uses consistent terminology | High |
| E3 | Does it contain specific data, cases, or verifiable claims? | Check for numbers, cases, citations | High |
| E4 | Are entity relationships clearly stated (e.g., "A is a superset of B")? | Check if logical relationships between concepts are clear | Medium |
| E5 | Is there a structured question-answer pair format? | Check if the article covers questions readers might ask | Medium |
Scoring Criteria: 20 points per item, maximum 100 points.
Detection Goal: Confirm whether the article has signals that make it trustworthy to AI
| Item ID | Check Item | Method | Weight |
|---|---|---|---|
| A1 | Is there a clear author byline? | Check if the article has the author's name or pen name | High |
| A2 | Is there an author bio or background description? | Check if there is an author introduction | Medium |
| A3 | Are authoritative sources cited (academic papers/official docs, etc.)? | Check references for authoritative sources | High |
| A4 | Are data points accompanied by verifiable source links? | Check if data includes source URLs | High |
| A5 | Are publication and update dates clearly marked? | Check if dates are displayed on the article | Medium |
| A6 | Is there a link to an "About/Contact" page? | Check if there is a link to author/institution introduction | Low |
Scoring Criteria: A1-A5 = 16.7 points each, A6 = 16.5 points, maximum 100 points.
Theoretical Basis: Content with high trustworthiness is 3.7 times more likely to be cited by AI.
Detection Goal: Confirm whether the article is easy for AI to cite
| Item ID | Check Item | Method | Weight |
|---|---|---|---|
| R1 | Is there a quotable "golden line" or core judgment statement? | Check for sentences that can be cited independently | High |
| R2 | Is the core argument summarized within the first 40-60 words? | Check if there is a conclusive statement at the beginning | High |
| R3 | Are there "citable statistics" available? | Check for hard data with sources | High |
| R4 | Does the article title contain technical keywords? | Check if the title has searchable technical terms | Medium |
| R5 | Is there a clear question-answer structure? | Check if the article answers a clear question | Medium |
Scoring Criteria: 20 points per item, maximum 100 points.
Detection Goal: Confirm whether information in the article is consistent across sources
| Item ID | Check Item | Method | Weight |
|---|---|---|---|
| X1 | Is the core concept self-consistent within the article? | Check if the same concept is expressed consistently across sections | High |
| X2 | Are data points consistent within the article? | Check for contradictions in data | High |
| X3 | Are core viewpoints consistent with the author's other articles? | Compare with the author's other articles | Medium |
| X4 | Do cited sources corroborate each other? | Check if multiple sources point to the same conclusion | Medium |
Scoring Criteria: 25 points per item, maximum 100 points.
Theoretical Basis: Cross-source consistency is one of the core dimensions of GEO effectiveness measurement from an RAG perspective.
Detection Goal: Confirm whether the article has mechanisms to trigger reader interaction
| Item ID | Check Item | Method | Weight |
|---|---|---|---|
| D1 | Does the title contain engagement-triggering words (you/how/invite/seek/together)? | Manual check of the title | High |
| D2 | Does the opening contain counterintuitive judgments or cognitive conflict? | Check first 300 words for "counterintuitive" statements | High |
| D3 | Is there a specific, answerable engagement question at the end? | Check the end for open-ended questions | High |
| D4 | Is there a collaboration invitation or participation entry point? | Check for "invite readers to participate" content | Medium |
| D5 | Does the article contain "reflexive design" (demonstrating what it advocates)? | Check if the article demonstrates the method it teaches | Medium |
| D6 | Is there a "reward-style" question that encourages readers to verify? | Check for phrasing like "If you've also noticed... feel free to share" | Low |
Scoring Criteria: D1-D3 = 20 points each, D4-D6 = 13.3 points each, maximum 100 points.
Theoretical Basis: Based on historical data analysis of 28 articles β titles with engagement-triggering words have an average engagement depth of 5.42, compared to 3.72 without, a 31.4% increase.
Detection Goal: Confirm whether the article meets the requirements of standards such as T/CAPT 026β2026
| Item ID | Check Item | Method | Weight |
|---|---|---|---|
| G1 | Is the content authentic and traceable? | Check for fabricated data or unverifiable claims | High |
| G2 | Are facts and opinions clearly distinguished? | Check if personal opinions are presented as facts | High |
| G3 | Are sources clearly cited? | Check if data/citations have sources | High |
| G4 | Does it involve prohibited behaviors such as "corpus poisoning" or "answer hegemony"? | Check for intent to manipulate AI output | High |
| G5 | Is full-chain traceability supported? | Check for modification records or version notes | Medium |
Scoring Criteria: G1-G4 = 20 points each, G5 = 20 points, maximum 100 points.
Theoretical Basis: T/CAPT 026β2026 "Generative Engine Optimization (GEO) β Trustworthy Information Dissemination and Information Ecology Governance Specification" requires content to be authentic, traceable, fully chain-operable, and delivered with white-box transparency.
Detection Goal: Confirm whether the article adopts content formats preferred by AI
| Item ID | Check Item | Method | Weight |
|---|---|---|---|
| F1 | Does it have the "one-sentence answer + table + FAQ" structure? | Check if the article contains all three elements | High |
| F2 | Is there a 40-60 word conclusion placed at the beginning of each section? | Check if each section starts with a summary sentence | Medium |
| F3 | Are visual elements like Mermaid diagrams used? | Check for flowcharts/architecture diagrams | Medium |
| F4 | Is there multimodal content (text + images + tables)? | Check if content formats are diverse | Medium |
| F5 | Does it contain an llms.txt entry (if applicable)? | Check if the site root has llms.txt |
|
| Low |
Scoring Criteria: 20 points per item, maximum 100 points.
Detection Goal: Confirm whether the article has "communication units" that can be remembered and cited by AI
| Item ID | Check Item | Method | Weight |
|---|---|---|---|
| M1 | Is there an independently quotable golden line (counterintuitive/shareable)? | Check for sentences like "The article with the lowest readership had the highest engagement depth" | High |
| M2 | Is there a central metaphor running through the entire article? | Check if there is a concrete metaphor anchoring the whole article | High |
| M3 | Is there a reusable formula for generating golden lines? | Check if the "construction method" of the golden line is made explicit | Medium |
| M4 | Is there a memorable one-sentence conclusion? | Check the end for a "one-sentence wrap-up" | High |
| M5 | Does it have a complete "source metaphor + metaphor chain" structure? | Check if multiple sub-metaphors are derived from a "source metaphor" | Medium |
Scoring Criteria: 20 points per item, maximum 100 points.
Theoretical Basis: Based on the "Metaphor & Quote Formula" module design from the published article "Skill Doesn't Need to Be Skill-ified." Sentences that can be remembered and shared are the core material for AI citation.
| Step | Action | Notes |
|---|---|---|
| 1 | Open the full page of the article to be detected | Ensure the page is fully loaded |
| 2 | Open the page source code (Ctrl+U or right-click "View Page Source") | Used for structured data detection |
| 3 | Prepare a detection record form | Can use Excel or pen and paper |
It is recommended to detect modules in the following order, recording scores after each module:
Total Estimated Time: Approximately 80-90 minutes per article.
ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
β GEO DETECTION REPORT β
β Article Title: [Title] β
β Detection Date: [YYYY-MM-DD] β
β Detected By: [Name] β
ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ€
β I. Comprehensive GEO Score: [X]/100 β
β β
β II. Module Score Breakdown β
β Module 1 Technical Crawlability: [X]/100 ββββββββββ [Rating] β
β Module 2 Structured Data Markup: [X]/100 ββββββββββ [Rating] β
β Module 3 Content Structure: [X]/100 ββββββββββ [Rating] β
β Module 4 Semantic Clarity: [X]/100 ββββββββββ [Rating] β
β Module 5 Trustworthiness: [X]/100 ββββββββββ [Rating] β
β Module 6 Citation Readiness: [X]/100 ββββββββββ [Rating]β
β Module 7 Cross-Source Consistency: [X]/100 ββββββββββ [Rating]β
β Module 8 Engagement Depth: [X]/100 ββββββββββ [Rating] β
β Module 9 GEO Compliance: [X]/100 ββββββββββ [Rating] β
β Module 10 AI-Friendly Format: [X]/100 ββββββββββ [Rating]β
β Module 11 Metaphor & Quote Readiness: [X]/100 ββββββββββ [Rating]β
ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ€
β III. Low-Score Diagnostics & Optimization Recommendations β
β [Item ID]: [Current Status] β [Recommended Action] β
β [Item ID]: [Current Status] β [Recommended Action] β
β ... β
ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ€
β IV. Optimization Priority Ranking β
β π΄ High Priority (highest impact, implement immediately) β
β π‘ Medium Priority (implement in near term) β
β π’ Low Priority (can optimize opportunistically) β
ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ€
β V. Expected Post-Optimization Results β
β Expected AI Citation Rate Increase: [X]% β
β Expected Engagement Depth Increase: [X]% β
ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
| Score Range | Grade | Status Description |
|---|---|---|
| 90-100 | A | Excellent β very high AI citation friendliness, maintain current standards |
| 70-89 | B | Good β minor optimization opportunities, targeted improvements recommended |
| 50-69 | C | Average β clear gaps present, systematic optimization recommended |
| 30-49 | D | Below Average β multiple dimensions require improvement |
| 0-29 | E | Poor β article structure should be re-evaluated |
This protocol supports on-demand module selection. Recommended configurations for different scenarios are as follows:
| Scenario | Recommended Modules | Notes |
|---|---|---|
| Quick Screening | ||
| Modules 1, 3, 5, 6 | 15 minutes for basic assessment | |
| Deep Optimization | ||
| All 11 modules | 90 minutes full assessment | |
| Technical Articles | ||
| Modules 1, 2, 3, 4, 5, 10 | Focus on structure and trustworthiness | |
| Operational/Method Articles | ||
| Modules 3, 6, 8, 11 | Focus on engagement and citation readiness | |
| GEO Compliance Audit | ||
| Modules 1, 5, 9 | Focus on compliance | |
| Pre-Publication Check | ||
| Modules 1, 2, 3, 4, 5, 6, 8 | Final check before publishing |
| Item | Issue When Score is Low | Recommended Action |
|---|---|---|
| T1 | Article requires login to view | Set article to "Public" in CSDN settings |
| T2 | Content is in image/PDF format | Ensure body text is HTML text format |
| T3 | No <title> or unclear title |
|
| Set article title in CSDN editor | ||
| T4 | No <meta description> |
|
| Fill in article summary in CSDN editor | ||
| T5 | robots.txt blocks AI crawlers | Modify robots.txt at site root, add: Allow: GPTBot , Allow: ClaudeBot |
| Item | Issue When Score is Low | Recommended Action |
|---|---|---|
| S1-S5 | Missing JSON-LD structured data | Add <script type="application/ld+json"> to the page, including Article , Author , FAQPage Schemas |
JSON-LD Template Example:
{
"@context": "https://schema.org",
"@type": "Article",
"headline": "Article Title",
"description": "Article Summary",
"author": {
"@type": "Person",
"name": "Author Name"
},
"datePublished": "2026-08-22",
"dateModified": "2026-08-22"
}
| Item | Issue When Score is Low | Recommended Action |
|---|---|---|
| C1 | No clear H1 heading | Ensure the article has a clear # title |
| C2 | No H2/H3 hierarchy | Use ## and ### for sub-sections |
| C3 | No summary or summary too long | Add a 150-200 word summary at the beginning |
| C4 | No lists/tables | Insert <ul> /<ol> or tables where appropriate |
| C5 | Section content insufficient | Add at least 300 words under each H2 |
| C6 | No section summary | Add 1-2 summary sentences at the end of each section |
| Item | Issue When Score is Low | Recommended Action |
|---|---|---|
| E1 | Core concept not defined | Provide a clear definition at first appearance |
| E2 | Terminology inconsistent | Unify terminology throughout the article |
| E3 | Missing specific data/cases | Add verifiable data or cases |
| E4 | Entity relationships unclear | Use clear phrasing like "A is the parent of B..." or "A consists of B..." |
| E5 | No question-answer structure | Embed FAQ or question-driven sections in the article |
| Item | Issue When Score is Low | Recommended Action |
|---|---|---|
| A1 | No author byline | Add author name at the beginning or end of the article |
| A2 | No author bio | Add 2-3 sentences of author introduction at the end |
| A3 | No authoritative citations | Cite authoritative sources such as academic papers, official documentation |
| A4 | Data without source links | Add a verifiable URL for each data point |
| A5 | No publication date | Mark publication and update dates at the beginning or end of the article |
| A6 | No "About/Contact" link | Add an "About the Author" or contact link in the author bio section |
| Item | Issue When Score is Low | Recommended Action |
|---|---|---|
| R1 | No "golden line" | Distill core judgment statements and place them prominently |
| R2 | No core conclusion in the opening | Summarize the core argument within the first 40-60 words |
| R3 | No citable statistics | Supplement with hard data and sources |
| R4 | Title lacks technical keywords | Add searchable technical terms to the title |
| R5 | No question-answer structure | Embed clear Q&A format in the article |
| Item | Issue When Score is Low | Recommended Action |
|---|---|---|
| X1 | Same concept expressed inconsistently | Unify terminology throughout the article |
| X2 | Data contradictions | Check and unify all data points |
| X3 | Contradicts the author's other articles | Check consistency across series articles |
| X4 | Cited sources contradict each other | Verify cited sources and ensure they corroborate |
| Item | Issue When Score is Low | Recommended Action |
|---|---|---|
| D1 | Title lacks engagement-triggering words | Add "you/how/invite/seek/together" to the title |
| D2 | Opening lacks cognitive conflict | Add counterintuitive statements within the first 300 words |
| D3 | End lacks engagement question | Add a specific, answerable question at the end |
| D4 | No collaboration invitation | Add "invite readers to participate" content at the end |
| D5 | No reflexive design | Have the article demonstrate the method it advocates |
| D6 | No reward-style question | Set questions like "If you've also noticed... feel free to share" |
| Item | Issue When Score is Low | Recommended Action |
|---|---|---|
| G1 | Content cannot be traced | Ensure all claims are verifiable |
| G2 | Facts and opinions are confused | Clearly distinguish "facts" from "my opinions" |
| G3 | No source citations | Add sources for all citations |
| G4 | Involves manipulation of AI output | Avoid prohibited behaviors such as "corpus poisoning" |
| G5 | No modification records | Add version number and modification date at the end |
| Item | Issue When Score is Low | Recommended Action |
|---|---|---|
| F1 | Missing "one-sentence answer + table + FAQ" structure | Add these three elements to the article |
| F2 | No conclusion at section beginnings | Add a 40-60 word conclusion at the start of each H2/H3 |
| F3 | No charts | Add Mermaid flowcharts or architecture diagrams |
| F4 | Content format is too uniform | Add images, tables, and other diverse formats |
| F5 | No llms.txt | Create llms.txt at the site root |
| Item | Issue When Score is Low | Recommended Action |
|---|---|---|
| M1 | No independently quotable golden line | Distill a counterintuitive core judgment statement |
| M2 | No central metaphor | Choose a concrete metaphor to run through the entire article |
| M3 | No golden line formula | Make the construction method of the golden line explicit |
| M4 | No one-sentence conclusion | Add a shareable concluding sentence at the end |
| M5 | No source metaphor chain | Derive multiple sub-metaphors from a "source metaphor" |
| Article Status | Detection Frequency | Notes |
|---|---|---|
| Before publishing new articles | Required for each article | Ensure basic GEO detection is completed before publishing |
| Already published articles | Once per quarter | Monitor AI citation rate changes and assess optimization needs |
| Engagement depth decline | Immediate detection | Check all dimensions when article "activity" declines |
| Series articles | Each article | Ensure consistency across series articles |
| Version | Date | Updates |
|---|---|---|
| v1.0 | 2026-08-22 | Initial version, includes 11 detection modules |