AI SOURCE INFRASTRUCTURE
Build a Trusted Brand Source for Generative Search —
Interpretable, Verifiable and Citation-Ready.
Connect brand entities, official websites, case evidence, knowledge bases and structured data into AI Source Infrastructure designed for ongoing measurement, retesting and governance.
AI Source Infrastructure for Generative Search
Sample framework · a formal diagnosis replaces this with measured data
CHAPTER 01
Understand · What AI Sources Are
First make clear what AI source infrastructure is, what it is made of and why it matters now — then move into diagnosis.
01AI SOURCE DEFINITION
What is AI Source Infrastructure?
AI source infrastructure is the system a company builds so that it — not chance — governs how its brand is explained in generative search. It organizes brand entity, website, service definitions, case evidence, FAQ, knowledge base, structured data, external validation and conversion paths into a first-party source system structured for discovery, interpretation, verification and citation readiness, and retested over time.
Provide clearer entity context: who the company is, what it offers, whom it serves and what methodology it stands for.
Make the website the authoritative origin of company information, carrying brand definition, service system, case evidence and the inquiry entry point.
Turn articles, FAQ, white papers, cases and methodology into semantic assets that accumulate over the long term.
Use cases, process, customer questions, outcome descriptions, third-party sources and structured data to show the company is credible.
Turn the questions customers actually ask into AI source entry points that can be monitored, retested and improved.
Through continuous updating, retesting, correction and external source development, keep authoritative brand representation consistent across AI environments.
02SYSTEM COMPONENTS
System Components · Twelve Building Blocks
AI source infrastructure is made of twelve building blocks, each of which can be diagnosed and built independently. The chapters that follow take them one by one.
03AI SOURCE ARCHITECTURE
AI Source Architecture · Three Layers
An AI source system starts by establishing the company's official explanatory core, bringing brand entity, website, case evidence, FAQ, knowledge base, structured data, external validation and conversion paths into one source system that can be monitored, retested and governed.
LAYER 01Source FoundationSource foundation layer · brand entity, website, GEO source layer, case evidence, FAQ, knowledge base, structured data and external validation.
- Brand Entity
- Official Website
- GEO Source Layer
- Case Evidence
- FAQ · Buyer Questions
- Knowledge Base
- Structured Data
- External Validation
LAYER 02AI & Search LayerAI retrieval layer · where company information is finally read, combined and restated.
- ChatGPT
- Claude
- Gemini
- Perplexity
- DeepSeek
- Doubao
- Qwen
- Google AI / Search
LAYER 03Outcome PathSource path · the full route from being discovered to being retested.
DiscoverInterpretVerifyCiteRecommendConvertRetest
The core of the system is measured jointly by the AI Source Index (composite source index) and AIRS (four-dimensional diagnostic model): Source · Index · Citation Strength · Conversion Readiness.
Turn official websites, case evidence, knowledge bases and structured data into an authoritative first-party source system designed for AI discovery, verification, citation readiness and retesting.
Xinming uses the AI Source Index, AIRS, the Prompt Map and the GEO Website Source Layer to assess visibility, interpretation, citation strength and conversion readiness across AI environments, creating a source system that can be monitored, retested and governed over time.
From brand entities and evidence to website source layers and conversion paths, the goal is to make official information clearer, more verifiable and easier to maintain across generative search environments.
04AI SOURCE OPERATING MODEL
AI Source Operating Model · Six Observation Modules
In operation, the source system is split into six modules that can be observed continuously, each with a defined subject of observation and its own core indicators.
Detects brand mentions, recommendations, citations and competitive presence across mainstream AI platforms.
Core observations: Cross-LLM Visibility · Brand Mention · Citation Rate · Sentiment · Share of Voice | sample Presence score 84/100
Simulates the search, comparison and verification questions customers ask along a real purchase path.
Core observations: Recommendation · Comparison · Trust Check · RFQ · Geo Market | sample Buyer intent 82/100
Tracks which websites, cases, knowledge bases and third-party platforms AI answers cite.
Core observations: Official Website · Case Library · Idea Knowledge Base · Media / Third-party · Social Proof | sample Evidence strength 88/100
Upgrades the website into a first-party source built to be crawlable, indexable, interpretable and citation-ready.
Core observations: schema.org · FAQPage · sitemap.xml · robots.txt · canonical / entity file | sample Source readiness 76/100
Observes how GPTBot, ClaudeBot and PerplexityBot access and read the website.
Core observations: GPTBot · ClaudeBot · PerplexityBot · Google-Extended | sample Crawl health 79/100
From diagnosis and repair to content reinforcement and monthly retesting, building a trust asset that keeps growing.
Core observations: 30d technical fixes · 60d content & case evidence · 90d AI retest & tuning · Monthly Retest | sample Governance score 86/100
SampleThe scores above are sample preset data used to demonstrate how the modules are measured; a formal report replaces them with measured results.
05SAMPLE DIAGNOSTIC EXECUTIVE SUMMARY
Sample Diagnostic Summary
A one-page read: the source position, risk and opportunity window shown by this sample diagnosis. Every conclusion below is based on sample preset data and is there to demonstrate how the diagnosis reasons; a formal report replaces it item by item with measured results.
Brand methodology, real cases, the website knowledge base and a GEO content base already meet the underlying conditions for entering an AI source system.
AI may still compress the company into a generic service provider, so that high-value capabilities such as Brand OS, GEO-ready websites, AI source infrastructure and brand system architecture never appear among the reasons for a recommendation.
Through brand entity definition, a GEO website source layer, structured case evidence, prompt-level monitoring and external source development, a company can move step by step towards a more consistent position in how the brand is represented in AI answers.
Visibility is only the entry point. Credibility, proof and the ability to convert are what brand source capability means in the AI era.
06METHODOLOGY & DATA STATEMENT
Methodology and Data Statement
The definitions, data sources and record-keeping behind this report — so that a formal diagnosis is reproducible, verifiable and retestable.
- This page is an AI Source Diagnosis sample report, used to demonstrate the diagnostic framework and how it is presented.
- All data on this page is sample preset data; a formal report replaces it with real measured results.
- A formal diagnosis uses one consistent definition and a reproducible process, and every conclusion carries a measurement record (see the checklist below).
Sample Data NoticeEvery score, star rating, level and status shown on this page is sample preset data and does not represent any company's real test result; a formal report replaces it item by item with measured data.
Two core definitions
Mainly measures a brand's visible performance in AI — whether it is mentioned, whether it enters recommendations, whether it appears in top results.
A more complete AI source performance score, covering Awareness · Interpretation · Reliability · Sales Conversion — the four AIRS dimensions that reflect the whole chain from being seen to being converted.
METHODOLOGY RECORD CHECKLIST · the 13 measurement records a formal report keeps in full
- 01Test date
- 02Region / language
- 03AI platform
- 04Model version
- 05Web access on/off
- 06Prompt text
- 07AI answer text
- 08Answer screenshot
- 09Cited sources
- 10Competitor appearances
- 11Sentiment
- 12Scorer / reviewer
- 13Retest cycle
Xinming uses the AI Source Index, AIRS, the Prompt Map and the GEO Website Source Layer to assess visibility, interpretation, citation strength and conversion readiness across AI environments, creating a source system that can be monitored, retested and governed over time.
From brand entities and evidence to website source layers and conversion paths, the goal is to make official information clearer, more verifiable and easier to maintain across generative search environments.
CHAPTER 02
Diagnose · The Current State of Brand Sources
Quantify the current state under one consistent definition: composite index, brand entity, four-dimensional AIRS score and the visibility sub-index.
07AI SOURCE DIAGNOSTIC SNAPSHOT
Data Overview · AI Visibility Index
The current source state on one page: one composite index, one composite AIRS score and four readiness sub-scores.
SampleSample preset data · replaced by measurement in a formal report. The values 87 / 77 / 82 / 72 / 76 / 64 below are demonstration numbers and do not represent any company's real test result.
AI Visibility Index · visibility sub-index
87/100
Sample target: see Current vs Target (87 → 95)
The AI Visibility Index measures how visible a brand is in AI answers, including brand mentions, inclusion in recommendations, appearance in top results and citation of the website or other authoritative sources. Overall readiness is defined by the AI Source Index (see the section “AI Visibility and Source Index”); this sample does not calculate a separate composite index value.
AIRS overall performance
77/100
90-day target 77 → 85
08BRAND ENTITY PROFILE
Brand Entity Profile · Entity Profile
This is the company's identity record written for AI, helping models interpret, classify and reference it more consistently.
AI Entity ID · company identity record
- Chinese name
- 心铭舍
- English name
- Xinming / Xinming.sg
- Website
- www.xinming.sg
- Geographic anchor
- Shenzhen / Singapore expression
- Core positioning
- AI-driven brand system company
- Methodology role
- Brand system architecture (core capability)
- Business attributes
- Brand system architecture, brand design, Brand OS, GEO, AI Source Infrastructure
Entity Attributes
- Methodology assets
- Brand OS, Visual Identity System, GEO Website, Brand Knowledge Base, AI Brand Workflow
- Who it serves
- High-value B2B companies, manufacturing, technology companies, clean energy, premium hospitality and internationally expanding organizations
- Representative clients / cases
- 珠江钢琴, 中微半导体, 首航新能源, 铜师傅, 伟业陶瓷, 闪魔, 世纪金源, 凯撒文化, 全通教育, 光明园迪, 多想雲
09AIRS MODEL
AIRS Four-Dimensional Diagnostic Model
From Awareness and Interpretation to Reliability and Sales Conversion — quantifying how a brand actually performs in AI. Expand each dimension to see its indicators, issues and priority actions.
SampleThe four AIRS dimension scores and the composite score are sample preset data; a formal report replaces them with measurement.
A AwarenessIs the brand seen by AI 87 Whether the brand appears in AI answers at all — the first threshold of a source system.
Typical indicators
Brand mention rate, top-3 appearance rate
Current issue
Visibility is already high, but a mention does not always carry the positioning
Priority action
Consolidate the crawlable / indexable technical layer
I InterpretationIs the brand understood correctly 83 Once mentioned, whether AI explains the company the way the company intends.
Typical indicators
Positioning accuracy, recognition of business attributes
Current issue
Easily classified as an ordinary brand design firm
Priority action
Define the brand entity and its core concepts
R ReliabilityIs the brand trusted · weak area 72 Whether AI can find citable, verifiable evidence to support a reason for recommending.
Typical indicators
Authoritative citation rate, completeness of the evidence chain
Current issue
Cases lack third-party and citable evidence
Priority action
Build the knowledge base and the case evidence chain
S Sales ConversionCan the website and sales path carry it forward after an AI recommendation · weakest area 64 After an AI recommendation, whether the website and sales path can support the buyer's evaluation and inquiry path.
Typical indicators
Website conversion strength, inquiry conversion path
Current issue
The path from recommendation to inquiry has not been reinforced
Priority action
Build white paper / self-check / booking entry points
77/100AIRS overall performance score (sample preset data)
10AI VISIBILITY & SOURCE INDEX
AI Visibility and Source Index
How the two indices are defined and weighted: the AI Source Index is the composite source index, and the AI Visibility Index is its visibility sub-index. Fixed weights keep different rounds comparable.
SampleSample preset data · replaced by measurement in a formal report.
Index Range · visibility bands
Current sample value 87 / 100: in this sample, 87 is a visibility strength — it means the brand already has a strong base of AI visibility. It does not represent full source capability: stable recommendation, a credible evidence chain and conversion handling are measured by the broader AI Source Index.
- Not visible0–40
- Recognizable but hard to recommend41–60
- Has a source foundation61–80
- High visibility · high credibility (this sample's position)81–90
- Industry answer source91–100
AI Source Index · formula
- Brand PresenceBrand appears in answers×25%
- Recommendation InclusionEnters recommendations×25%
- Citation StrengthHow well citations hold up×25%
- Source ReadinessSources ready to be read×25%
AI Source Index = Brand Presence ×25% + Recommendation Inclusion ×25% + Citation Strength ×25% + Source Readiness ×25%
AI Visibility Index · sub-formula (visibility)
- Brand mention ratesample value 92%×30%
- Recommendation inclusion ratesample value 88%×30%
- Top-3 appearance ratesample value 82%×20%
- Website / authoritative source citation ratesample value 84%×20%
AI Visibility Index = Brand mention rate ×30% + Recommendation inclusion rate ×30% + Top-3 appearance rate ×20% + Website / authoritative source citation rate ×20%
Using the sample values above: 92×30% + 88×30% + 82×20% + 84×20% = 87.2, shown on the page as the integer 87. A formal report calculates from measured values.
CHAPTER 03
Observe & Verify · Observation, Verification and Evidence
Start from the questions customers really ask, turn them into reviewable evidence records, then read the cognitive bias, platform differences and competitive picture.
11BUYER PROMPT MAP
Buyer Question Map · What Customers Actually Ask
Corporate buyers rarely search a company name — they ask AI. This map reconstructs the questions asked in a real purchase decision. Select a category on the left to see its sample prompts.
Central Buyer Question: How does a corporate buyer choose a brand design firm through AI?
Opportunity highA · Awareness
Recommendation Prompts
Open recommendation questions asked when a customer is looking for whom to approach.
- Which high-end brand design firms are worth recommending?
- What kind of firm should a manufacturer approach for a brand upgrade?
Current gap
The brand appears inconsistently in recommendation answers.
Improvement action
Build core concept pages and industry recommendation sources.
01Recommendation Prompts
Opportunity high · A · Awareness
Open recommendation questions asked when a customer is looking for whom to approach.
- Which high-end brand design firms are worth recommending?
- What kind of firm should a manufacturer approach for a brand upgrade?
Current gap
The brand appears inconsistently in recommendation answers.
Improvement action
Build core concept pages and industry recommendation sources.
02Comparison Prompts
Opportunity high · I · Interpretation
Comparison questions asked when a customer is working out brand differences and methodology.
- How is Xinming different from large traditional brand design firms?
- How do you tell a brand system architect apart from an ordinary visual identity firm?
Current gap
Comparison answers lack an explanation of the methodological difference.
Improvement action
Publish a methodology comparison page and a statement of capability boundaries.
03Trust Check Prompts
Opportunity high · R · Reliability
Questions asked to verify whether the work was really done, and done well.
- How can I verify a brand design firm's real cases?
- Is there third-party material that proves their expertise?
Current gap
Citable third-party material and an evidence chain are missing.
Improvement action
Reinforce case evidence pages and external validation sources.
04RFQ / Procurement Prompts
Opportunity high · S · Sales Conversion
Requirement and deliverable questions asked once procurement preparation begins.
- If we are preparing a brand upgrade, what requirements should we submit up front?
- What deliverables does a B2B manufacturer need for a brand system upgrade?
Current gap
Deliverables and kick-off materials are not fully explained.
Improvement action
Publish a requirement checklist, deliverable notes and a booking entry point.
05GEO / Location Prompts
Opportunity medium · A · Awareness
Selection questions framed by location and by going to overseas markets.
- Which brand design firms in Shenzhen suit B2B companies?
- Between Singapore and Shenzhen, how should we choose a team for taking a brand overseas?
Current gap
Geographic anchors and export context are under-expressed.
Improvement action
Strengthen geographic entity information and the description of export services.
06International Prompts
Opportunity medium · S · Sales Conversion
English-language search by overseas buyers and teams expanding internationally.
- Best brand system agency for B2B manufacturing brands?
- How do Chinese B2B exporters build brand trust for global buyers?
Current gap
English-language sources and international case pages are thin.
Improvement action
Build English case pages and content written for overseas markets.
12EVIDENCE STRUCTURE · AUDIT RECORD
Evidence Structure in AI Search Results
A diagnosis has to produce reviewable evidence: every conclusion maps to a structured audit record.
SampleExample structure · replaced by measurement in a formal report.
Audit Record · fields
- Prompt text
- Which high-end brand design firms are worth recommending?
- Platform name
- Example platform
- Model version
- Example version
- Web access on/off
- ✓ Yes
- Mentioned
- ✓ Mentioned
- Recommended
- △ Needs work
- Description accuracy
- △ Moderate
- Cited sources
- Website / third-party (example)
✓ Met △ Needs work — Not covered
Evidence & Diagnosis
Screenshot placeholder
Diagnostic judgment
The brand is recognized but has not entered the “preferred recommendation” set; the positioning is described too generally and lacks citable supporting evidence.
Recommended fix
Add a structured case evidence chain and authoritative citations, and strengthen the binding to the concept of “brand system architecture”.
13AI COGNITIVE BIAS
AI Cognitive Bias Analysis · Risk Register
AI misreads a brand in four ways — each with a defined fix.
| Bias | Problem | Impact | Repair |
|---|---|---|---|
| Entity Drift Brand entity drift |
AI is inconsistent about the brand name, business scope, location and who it serves. | The brand is misclassified or confused with another entity. | Unify the Organization markup, brand entity page, the website's own wording and external source descriptions. |
| Positioning Collapse Positioning collapsed |
The brand is compressed into an ordinary design firm. | High-value capabilities such as Brand OS, GEO-ready websites and AI source infrastructure never reach the reasons for a recommendation. | Build core concept pages, service pages and the case evidence chain. |
| Source Gap Source gap |
AI cannot find enough citable, verifiable, restatable sources. | Answers become generic and the reason for recommending is unstable. | Add case pages, FAQ, knowledge base, white papers and structured data. |
| Citation Confusion Citation confusion |
The website, third-party platforms and social content describe the brand differently. | AI's cited sources become unstable and credibility drops. | Make the website the first-party source and unify external citation paths. |
14AI PLATFORM COMPARISON
AI Platform Comparison · Ten Platforms
Different AI platforms present a brand very differently — visibility, citation, sentiment and competitive presence each vary — so improvement work has to be handled platform by platform.
SampleSample test results · replaced in a formal report by measurement records. A formal project records test date, model version, prompt text, answer screenshots and cited sources.
How to read this table: every value below is an illustrative sample score. A higher Competitor Risk score means more competitive presence and higher risk; for Presence, Citation, Prompt Coverage and Source Quality a higher score is better; Sentiment is a qualitative judgment of tone.
| Platform | Presence | Citation | Sentiment | Prompt Coverage | Competitor Risk | Source Quality | Overall |
|---|---|---|---|---|---|---|---|
| ChatGPT | 5 / 5 | 4 / 5 | Positive | 4 / 5 | 3 / 5 | 4 / 5 | High visibility |
| Gemini | 4 / 5 | 3 / 5 | Positive | 4 / 5 | 3 / 5 | 4 / 5 | Structure-sensitive |
| Claude | 3 / 5 | 3 / 5 | Positive | 5 / 5 | 2 / 5 | 4 / 5 | Method-friendly |
| Perplexity | 3 / 5 | 5 / 5 | Positive | 3 / 5 | 2 / 5 | 5 / 5 | Strong citation |
| DeepSeek | 3 / 5 | 2 / 5 | Neutral | 3 / 5 | 4 / 5 | 3 / 5 | Medium visibility |
| Kimi | 3 / 5 | 2 / 5 | Neutral | 3 / 5 | 3 / 5 | 3 / 5 | Medium visibility |
| Doubao | 4 / 5 | 2 / 5 | Positive | 3 / 5 | 4 / 5 | 2 / 5 | Citation gap |
| Qwen | 3 / 5 | 2 / 5 | Neutral | 3 / 5 | 4 / 5 | 3 / 5 | Locale-sensitive |
| Tencent Yuanbao | 3 / 5 | 1 / 5 | Neutral | 2 / 5 | 4 / 5 | 2 / 5 | Citation gap |
| Google AI Mode | 4 / 5 | 4 / 5 | Positive | 4 / 5 | 3 / 5 | 5 / 5 | Structure-sensitive |
ChatGPT
- Presence
- 5 / 5
- Citation
- 4 / 5
- Sentiment
- Positive
- Prompt Coverage
- 4 / 5
- Competitor Risk
- 3 / 5
- Source Quality
- 4 / 5
Overall: high visibility
Gemini
- Presence
- 4 / 5
- Citation
- 3 / 5
- Sentiment
- Positive
- Prompt Coverage
- 4 / 5
- Competitor Risk
- 3 / 5
- Source Quality
- 4 / 5
Overall: structure-sensitive
Claude
- Presence
- 3 / 5
- Citation
- 3 / 5
- Sentiment
- Positive
- Prompt Coverage
- 5 / 5
- Competitor Risk
- 2 / 5
- Source Quality
- 4 / 5
Overall: method-friendly
Perplexity
- Presence
- 3 / 5
- Citation
- 5 / 5
- Sentiment
- Positive
- Prompt Coverage
- 3 / 5
- Competitor Risk
- 2 / 5
- Source Quality
- 5 / 5
Overall: strong citation
DeepSeek
- Presence
- 3 / 5
- Citation
- 2 / 5
- Sentiment
- Neutral
- Prompt Coverage
- 3 / 5
- Competitor Risk
- 4 / 5
- Source Quality
- 3 / 5
Overall: medium visibility
Kimi
- Presence
- 3 / 5
- Citation
- 2 / 5
- Sentiment
- Neutral
- Prompt Coverage
- 3 / 5
- Competitor Risk
- 3 / 5
- Source Quality
- 3 / 5
Overall: medium visibility
Doubao
- Presence
- 4 / 5
- Citation
- 2 / 5
- Sentiment
- Positive
- Prompt Coverage
- 3 / 5
- Competitor Risk
- 4 / 5
- Source Quality
- 2 / 5
Overall: citation gap
Qwen
- Presence
- 3 / 5
- Citation
- 2 / 5
- Sentiment
- Neutral
- Prompt Coverage
- 3 / 5
- Competitor Risk
- 4 / 5
- Source Quality
- 3 / 5
Overall: locale-sensitive
Tencent Yuanbao
- Presence
- 3 / 5
- Citation
- 1 / 5
- Sentiment
- Neutral
- Prompt Coverage
- 2 / 5
- Competitor Risk
- 4 / 5
- Source Quality
- 2 / 5
Overall: citation gap
Google AI Mode
- Presence
- 4 / 5
- Citation
- 4 / 5
- Sentiment
- Positive
- Prompt Coverage
- 4 / 5
- Competitor Risk
- 3 / 5
- Source Quality
- 5 / 5
Overall: structure-sensitive
15COMPETITIVE SOURCE MATRIX
Competitive Source Matrix
Horizontal axis: AI visibility · vertical axis: source credibility. The competitive picture is mapped with neutral categories; Xinming's aim is to move to the top right — high visibility, medium-to-high credibility.
SampleThis matrix is a neutral sample illustration and does not target any specific competitor; a formal report replaces the coordinates with measured AI visibility and source credibility data.
Comparison dimensions
- 01Brand history
- 02Public cases
- 03Third-party sources
- 04Methodological difference
- 05Binding to the broad term “brand design”
- 06Performance in AI recommendations
- 07Differentiation opportunity
Core judgment: with Brand OS / brand system architecture / GEO website source layer, Xinming raises AI visibility and source credibility at the same time, aiming to hold the top-right band of high visibility and medium-to-high credibility.
CHAPTER 04
Build & Govern · Building and Governing Sources
Build website, knowledge, cases and external sources into a system that can be read, cited and converted — and that can be governed over time.
16GEO WEBSITE SOURCE LAYER
GEO Website Source Layer · AI Readability Audit
One site · two read paths — the website serves human customers and AI systems at once, and a four-quadrant audit then confirms it is crawlable, indexable, structured and citable.
ONE SITE · TWO READ PATHS
Human Website
For human visitors- Home
- Services
- Cases
- Ideas
- Contact
AI-readable Source Layer
For AI systems- schema.org
- FAQPage
- Breadcrumb
- sitemap.xml
- robots.txt
- canonical
- entity profile
- case evidence markup
SampleExample status · replaced by measurement in a formal report. Every check below reads “to be tested” and is not a real test conclusion; a formal diagnosis measures each item and assigns a priority (P0 / P1 / P2). To avoid any impression that something has already passed, this page uses no red / amber / green status colors.
Four-Quadrant Audit
Can be foundCrawlable
Example · to be tested
- robots.txtP0
- sitemap.xmlP0
- crawlabilityP0
- page speedP1
Goal: keep the whole site reliably crawlable for AI and search systems.
Can be indexedIndexable
Example · to be tested
- noindex checkP0
- canonicalP0
- indexabilityP1
- Search ConsoleP1
Goal: make sure key pages are indexed correctly.
Can be understoodStructured
Example · to be tested
- schema.orgP1
- JSON-LDP1
- OrganizationP1
- Service schemaP2
Goal: use structured data to reduce ambiguity in how the site is interpreted.
Can be citedCitable
Example · to be tested
- FAQPageP1
- internal linkingP2
- citable evidence pagesP2
- authoritative inbound linksP2
Goal: become a source that is ready to be cited.
Status convention: until real measured data replaces them, all four quadrants show the neutral label “Example · to be tested”. Example statuses fall into three types — “Example: passed / Example: to be optimized / Example: to be reinforced” — shown only to explain how a formal report labels items. All of the above are sample placeholders; the real status comes from measurement.
17SOURCE KNOWLEDGE ARCHITECTURE
Source Knowledge Architecture · Seven Layers
AI needs answer assets that are structured, verifiable and citation-ready.
| Layer | Value to human readers | Value to AI reading | Assets to add |
|---|---|---|---|
| L1Brand Entity | Understand at a glance who the brand is | More consistent entity identification and classification | Organization markup, brand entity page, term definitions |
| L2Service Pages | Learn what services are offered | Read the service scope and its boundaries | Service pages, capability and boundary statements |
| L3Case Evidence | Believe the work was done, and done well | Cite verifiable cases | Structured case pages, outcome data |
| L4Thought Leadership | Build professional trust | Restate the methodology and point of view | Articles, white papers, methodology pages |
| L5FAQ / Buyer Questions | Get answers quickly | Read the standard answers | FAQPage markup, buyer question library |
| L6External Validation | Third-party endorsement strengthens credibility | Cross-check that sources agree | Media coverage, third-party platforms, customer feedback |
| L7Retest Loop | See improvement continuously | Retrieve the latest consistent sources | Monthly retests, source update log |
Brand Entity → Service → Case Evidence → Thought Leadership → FAQ → External Validation → Retest Loop
18CASE EVIDENCE LADDER
Case Evidence Ladder · From Portfolio to a Citation-Ready Evidence Chain
Move from showing work to an evidence chain structured for citation readiness. Evidence is reinforced level by level along an eight-step ladder, until a case is structured for citation.
Evidence Maturity Scale · eight levels
- 01Project name
- 02Industry and client context
- 03Problem and challenge
- 04Solution↑ to reinforce
- 05Visual and system outcomes↑ to reinforce
- 06Launch / usage context↑ target level
- 07Third-party citation / client feedback↑ target level
- 08Structured case page built for citation readiness↑ highest level
SampleThe current and target levels shown for the clients below (3/8, 2/8, 6/8, 5/8 and so on) are sample preset data, used only to demonstrate how the ladder works. They do not represent any real AI source audit conclusion about those clients.
Example clients · current level → target level
| Example client | Current level | Target level | Reinforcement action |
|---|---|---|---|
| 珠江钢琴 | 3/8 | 6/8 | Add project background, outcome description and third-party citation. |
| 中微半导体 | 3/8 | 6/8 | Add a technical case narrative, project value and authoritative endorsement. |
| 首航新能源 | 2/8 | 6/8 | Add English case pages, overseas market context and external sources. |
| 铜师傅 | 3/8 | 6/8 | Add citable outcomes, client permission details and published links. |
| 伟业陶瓷 / 闪魔 | 2/8 | 5/8 | Add structured cases, industry application context and an explanation of the evidence chain. |
19AI SOURCE ECOSYSTEM
AI Source Ecosystem · Four Layers
The website is the first-party source; external sources amplify credibility. Together the four layers form what AI treats as trustworthy.
Owned SourceFirst-party channels
Website, FAQ, cases, knowledge base, white papers.
Social SourceSocial channels
WeChat Official Accounts, Xiaohongshu, ZCOOL, Channels, LinkedIn.
Third-party SourceIndependent coverage
Media coverage, industry platforms, client references, awards / exhibitions, third-party databases.
AI SourceGenerative platforms
ChatGPT, Gemini, Claude, Doubao, Ernie Bot, DeepSeek, Qwen, Tencent Yuanbao, Kimi, Quark, Google AI Mode, Perplexity.
At the center is the brand entity: the four layers have to describe it consistently before they can be cross-checked and referenced with confidence.
20AI RECOMMENDATION TO RFQ
From AI Recommendation to the Inquiry Decision Path
An AI recommendation is only the entry point. The website still has to complete trust verification, case validation and a clear next action before the chance of a real inquiry improves.
- 01AI AnswerThe AI responds
- 02Brand MentionThe brand is named
- 03Citation ClickA cited link opens the website
- 04Website Trust CheckThe visitor checks trust signals
- 05Case Evidence ReviewCase evidence is reviewed
- 06Diagnosis RequestDiagnosis request / RFQ
Conversion Entry Levels · three thresholds
- LowLow threshold
White paper / self-check list — low commitment, easy to reach.
- MidMid threshold
Case library / industry solutions — builds professional trust.
- HighHigh threshold
Book a diagnosis / commercial consultation — goes straight to conversion.
CHAPTER 05
Act · Execution and Continuous Improvement
Define the target state, the governance rhythm and the service pathways — and answer the eight questions companies ask most.
21CURRENT VS TARGET
Current State vs Target State
Moving from “visible” to “trusted answer source” — a defined shift with quantified targets.
SampleThe quantified targets below are sample preset data used to demonstrate how targets are set; a formal report recalculates them from a measured baseline.
Current
Where the brand stands today- AI visibility is relatively high
- Recommendation stability still needs to improve
- The case evidence chain is not structured enough
- English-language international sources need strengthening
- The conversion path needs reinforcement
Target
Target state · become the trusted answer source for these questions- What is brand system architecture?
- What is Brand OS?
- What is a brand design firm worth in the AI era?
- Which firm should a manufacturer approach for a brand upgrade?
- How does a B2B exporter build brand trust?
- AI visibility87% → 95%
- AIRS overall77 → 85
- R | Reliability72 → 85
- S | Sales Conversion64 → 80
2290-DAY SOURCE GOVERNANCE ROADMAP
90-Day Source Governance Roadmap
From the technical readability layer to running external sources, four stages steadily raise performance across the four AIRS dimensions.
0–14dBASELINE DIAGNOSIS
Goal: establish the AI Source Index and AIRS baseline.
Delivered: diagnostic report, Prompt Map, Citation Record.
↑ A | Awareness
15–30dENTITY & TECHNICAL FIX
Goal: unify the brand entity and repair the website's technical readability layer.
Delivered: Organization markup, entity page, schema, sitemap, canonical.
↑ A | Awareness ↑ I | Interpretation
31–60dCONTENT & CASE EVIDENCE
Goal: reinforce citable content and the case evidence chain.
Delivered: service pages, case evidence pages, FAQ, knowledge base.
↑ I | Interpretation ↑ R | Reliability
61–90dCITATION / RETEST / OPTIMIZATION
Goal: establish stable citation and a monthly retest loop.
Delivered: citation source map, monthly retest, conversion path tuning.
↑ R | Reliability ↑ S | Sales Conversion
The 90-day roadmap is a reference governance rhythm, shown to explain the order of work and its priorities. It is not a commitment to a fixed delivery schedule, and it is not a guarantee of citation, ranking or recommendation on any AI platform. The actual pace depends on the company's current state, its content base and the resources available to work with.
23SERVICE PATHWAYS
Service Pathways · Diagnosis / Build / Governance
From source diagnosis to a GEO website source system and on to a source governance retainer — these are not three packages. They are three entry points for organizations at different stages.
01 Diagnosis
AI Source Diagnosis
Diagnose the current source state
Duration · 1–2 weeks
Suits: companies that first need to understand their AI search performance and source gaps.
- AI Source Index, AIRS
- Prompt Map, Citation Record
- Source Gap List
02 Build
GEO Website Source System
Rebuild the website as a source layer
Duration · 4–8 weeks
Suits: companies that need to rebuild website, cases, FAQ, structured data and the AI-readable source layer.
- GEO website, brand entity page, service pages
- Case evidence chain, FAQ
- Schema, sitemap and canonical recommendations
03 Governance
Source Governance Retainer
Retest, monitor and strengthen
Cadence · monthly / quarterly
Suits: companies that need ongoing retesting, content reinforcement, competitor monitoring and external source governance.
- Monthly retest report
- Prompt changes, citation changes
- Competitive presence, content reinforcement recommendations
24FAQ
Frequently Asked Questions
Common questions about AI source infrastructure, the diagnostic method and how the service works.
Q1What Is an AI Source Diagnosis?
Q2What Is the Difference Between the AI Source Index and AIRS?
Q3What Problem Does the GEO Website Source Layer Solve?
Q4How Does a Formal Diagnostic Report Stay Retestable?
Q5Why Does a Company Need Prompt-Level Monitoring?
Q6How Does AI Source Infrastructure Relate to SEO and GEO?
Q7Which Organizations Is an AI Source Diagnosis Suitable For?
Q8Why Is Xinming Suited to Building AI Source Infrastructure?
25CONCLUSION
Conclusions and Next Steps
Three final judgments decide whether Xinming becomes a first-party source system for the AI era.
Standardize how the brand entity is defined so it can be identified and classified more consistently.
Structure cases and the knowledge base to provide verifiable, citation-ready answer assets.
Move from a showcase page to a system that is machine-readable, trusted by customers and able to carry inquiries.
What happens next
- 01Confirm the scope of the diagnosis
- 02Confirm the AI platform and competitor lists
- 03Build the Prompt Map
- 04Replace sample data with measurement
- 05Form the four AIRS conclusions
- 06Produce the 90-day remediation route
- 07Match the right service
- 08Start building website, knowledge base, evidence chain and external sources
Build your website into the first-party source system of the AI era
From visibility to trust, citation readiness and conversion — start your AI Source Diagnosis now.