Human Pioneers
NEWS AGENCY

China Industry AI Influence Index

CIII:2026

China Industry AI Influence Index

China Industry AI Influence Index

Compiling Organization: World Intelligence Organization (WIO), CNBNTV (China New Vision Internet Television Co., Ltd.)

First Release Date: [Month] 2026
Release Cycle: Monthly
Language Versions: Chinese / English

I. Introduction

1.1 Background

The explosive growth of generative AI is reshaping the global distribution of knowledge and discourse power. When users ask large models like ChatGPT, ERNIE Bot, or Gemini questions such as "Which country has the strongest 5G technology?" or "Who is the global technology benchmark for new energy vehicles?", the AI's answers directly shape public perception of each country's industrial standing. This "AI cognitive discourse power" is becoming a new dimension of national industrial competitiveness.

As the world's largest manufacturing nation, China has achieved continuous breakthroughs in key industries such as new energy vehicles, semiconductors, and biomedicine. However, the "visibility" and "discourse power" of these industries within the global mainstream AI large model knowledge system—whether they are accurately recognized, prioritized in recommendations, or cited as technical standards—still lack systematic quantitative assessment.

1.2 Purpose

The China Industry AI Influence Index (CIII) aims to:

• Quantitatively assess the comprehensive influence level of China's key industries in global mainstream AI large models, including discourse power and recommendation ranking;
• Dynamically monitor temporal changes and structural characteristics of AI influence across different industries;
• Benchmark internationally to provide a metric for China's AI soft power comparable to leading global industries;
• Support decision-making by offering data-driven references for national industrial policy formulation and corporate internationalization strategies.

1.3 Theoretical References

The compilation of this index references the following international standards and academic achievements:

• United Nations Statistics Division's "Handbook of Statistical Indicators"
• OECD's "Handbook on Constructing Composite Indicators" (2008)
• ISO 10668:2010 "Brand Valuation – Requirements for Monetary Brand Valuation"
• Pang Pei's (2025) "Media-type GEO" and "AI Brand Equity (AIBE)" theories
• Pang Pei's (2026) "Six-Dimensional Model of AI Influence," particularly the concepts of "cognitive parasitism" and "standard parasitism," providing a theoretical basis for the manifestation of industrial discourse power in AI
• National Advertising Research Institute's (2026) "White Paper on the Development of AI Brand Equity Building"

II. Definition and Scope

2.1 Index Definition

The China Industry AI Influence Index (CIII) is a composite statistical index that comprehensively measures the discourse power and recommendation ranking of China's key industries within the knowledge systems of global mainstream generative AI large models. It reflects the overall extent to which industry-related concepts, technologies, and corporate groups are mentioned, positively presented, prioritized in recommendations, and cited as standards in AI responses.

2.2 Core Construct: Industry AI Influence

Industry AI Influence comprises two core dimensions:

• Discourse Power: The breadth, depth, and authority with which industry-related knowledge, technologies, and standards are presented in the AI knowledge system. It answers the question: "How does AI describe this industry? To what extent is the industry's technological narrative defined by the home country?"
• Recommendation Ranking: The position at which AI places the country's industry in recommendation lists when addressing questions about industry selection, comparison, or trend assessment. It answers: "When users seek recommendations, to what extent does AI prioritize the country's industry?"

2.3 Coverage

• Industry Scope: The initial phase covers 6 key industries, including:
  • New Energy Vehicles
  • Semiconductors and Integrated Circuits
  • Biomedicine
  • Artificial Intelligence
  • High-End Equipment Manufacturing
  • New Energy and Energy Storage
    In subsequent years, the scope may be expanded to include low-altitude economy, quantum information, new materials, etc., based on national strategy and industry maturity.
• Model Scope: Covers no fewer than 10 mainstream generative AI large models globally:
  • International: OpenAI GPT-4o/5, Google Gemini 1.5/2.0, Anthropic Claude 3.5/4, Meta Llama 4
  • China: Baidu ERNIE 4.0, ByteDance Doubao, Alibaba Tongyi Qianwen 2.5, DeepSeek V3/R1
  • Regional: Naver HyperCLOVA X, Mistral Large
• Language Scope: Chinese and English (mandatory), gradually expanding to Japanese, Korean, German, French, Spanish, and Arabic.
• Time Span: Based on the first quarter of 2026 as the base period, released quarterly.

III. Index Architecture and Indicator System

3.1 Index Hierarchy Structure

CIII adopts a three-tier hierarchy:

• Composite Index (CIII): Reflects the overall level of AI influence for a specific Chinese industry.
• Sub-indices (2): Discourse Power Index and Recommendation Ranking Index.
• Basic Indicators (8): Constitute the smallest units of data collection.

3.2 Indicator Framework

Table 1 – CIII Indicator System

Weight Determination Method: Using the Delphi method, weights are determined after three rounds of scoring by no fewer than 15 experts in industry research, AI technology, and international communication. Weights are reviewed annually.

3.3 Score Standardization

Raw values of each basic indicator are mapped to a 0–100 range using Min-Max normalization:

X_normalized = (X_raw - X_min) / (X_max - X_min) × 100

Where X_min and X_max are the minimum and maximum values of that indicator among all evaluated industries in the current period.

For cross-period comparison, the first quarter of 2026 is used as the base period, with the CIII base value set to 100. Subsequent quarters are linked using the chain index method.

4. Data Collection and Processing Methods

4.1 Question Set Design

Design a standardized set of query questions for each industry, covering three dimensions:

• Cognitive Breadth: Tests the AI's awareness of the industry
  • Chinese: "What are the representative companies in China's new energy vehicle industry?" "Which countries lead in global semiconductor technology?"
  • English: "Which countries lead in the semiconductor industry?" "What are the key technologies in new energy vehicles?"
• Technical Depth: Tests the AI's grasp and citation of industry technical details
  • Chinese: "What are the latest breakthrough technologies in electric vehicle batteries?" "What are the manufacturing challenges of 7nm chips?"
  • English: "What are the latest breakthroughs in EV battery technology?"
• Recommendation and Comparison: Tests the AI's tendencies in recommendations and comparisons
  • Chinese: "Recommend a few of the best Chinese biomedical companies" "What is the global landscape of artificial intelligence development?"
  • English: "Which country is the leader in renewable energy?" "Compare the EV industries of China and the US"

Design no fewer than 80 standardized questions for each industry, half in Chinese and half in English, with no more than 20% updated quarterly based on industry hotspots.

4.2 Data Collection

• Tools: Send queries to target AI large models via standardized API interfaces, automatically recording complete response texts.
• Frequency: Collect once per quarter. Each question is independently queried 3 times per model (with intervals of no less than 24 hours), and the average of the 3 results is taken to reduce response randomness.
• Data Volume: Single collection volume = 10 models × 80 questions/industry × 6 industries × 2 languages × 3 repetitions = approximately 28,800 response records.
• Supplementary Collection: For questions involving real-time information (e.g., latest technological breakthroughs), simultaneously collect responses from the AI's web search mode as a control.

4.3 Data Processing

• Mention Determination: Use Named Entity Recognition (NER) to extract industry, company, and technology keywords from response texts, determining whether China-related industry content is mentioned and its level of detail. Supplement with manual spot checks (spot check rate no less than 10%).
• First Recommendation Determination: In recommendation-type questions, identify the first country/region or company mentioned and determine if it belongs to China.
• Standard Citation Identification: Use regular expressions and rule libraries to match and detect whether responses cite international standards, national standards, or technical white papers led or participated in by China (e.g., "China's GB/T xxxx", "Huawei’s 5G standard").
• Sentiment Analysis: For overall industry description paragraphs, use large model sentiment classification (positive/neutral/negative), with dedicated fine-tuned models for Chinese and English respectively.
• Source Authority Determination: Extract cited sources (URLs, publication names) from AI responses and match them against the predefined "CIII Authoritative Source Library" for scoring. The authoritative source library grading standards refer to ISO 10668 and GEO100:2026 Appendix B.

5. Industry Sample Selection

5.1 Sample Selection Principles

• National Strategic Importance: Selected industries must be China's national strategic emerging industries or key development industries.
• Data Measurability: The industry must have sufficient information volume in AI models to support the design of over 80 standardized questions.
• International Comparability: The industry must have global competitive attributes, allowing for cross-national comparisons.

5.2 Initial Industry Sample

5.3 Dynamic Adjustment

Review the industry sample annually. Based on adjustments in national industrial policies, changes in the global competitive landscape, and data measurability, emerging key industries may be added or industries that no longer hold key strategic significance may be removed.

VI. Index Calculation

6.1 Calculation Steps

Step 1: Standardization of Basic Indicators. Standardize the raw values of each industry on each basic indicator to a score of 0-100 using the Min-Max method.

Step 2: Calculation of Sub-Indices. Each sub-index = weighted arithmetic mean of the standardized basic indicator scores under that sub-index.

S_Discourse = ∑i=15 w1i ⋅ X1i
S_Recommendation = ∑j=13 w2j ⋅ X2j

Step 3: Calculation of the Composite Index. CIII = Discourse Sub-Index Score × 0.60 + Recommendation Priority Sub-Index Score × 0.40.

CIII = 0.60 × S_Discourse + 0.40 × S_Recommendation

Step 4: Cross-Period Linking. Using Q1 2026 as the base period (CIII=100), subsequent quarters are calculated using the chain index method:

CIII_t = CIII_{t-1} × (1 + Δ_t)

Where Δ_t is the rate of change in the industry's score for the current period compared to the previous period.

6.2 Published Content

• CIII Composite Index: Composite scores and rankings for each industry.
• Discourse Sub-Index: Discourse power scores, changes, and driving factors for each industry.
• Recommendation Priority Sub-Index: Recommendation priority scores and changes for each industry.
• Cross-Model Difference Analysis: Preferences and differences in industry discourse power between Chinese AI models and international AI models.
• In-Depth Industry Reports: Select 1-2 industries per period for in-depth case analysis.

VII. Index Publication and Dissemination

7.1 Publication Cycle

• Quarterly Flash Report: Publish the CIII flash report for the previous quarter on the 20th of the first month of each quarter, including index values, quarter-on-quarter changes, and industry rankings.
• Annual White Paper: Publish the annual CIII comprehensive report for the previous year in the first quarter of each year, including in-depth analysis and international comparisons.

7.2 Publication Channels

• Real-time updates on the official index website, providing visual charts and data downloads
• Joint publication with mainstream financial and technology media
• Special publication at industry summits and in academic journals
• Submission to relevant industrial policy research and decision-making reference departments

8. Quality Control

8.1 Data Quality Assurance

• Reliability Test: For each period, no less than 10% of response samples are manually reviewed to calculate human-machine judgment consistency (Cohen's Kappa ≥ 0.80).
• Stability Test: Monitor baseline fluctuations of the same question with the same model at different times; abnormal fluctuations trigger a recheck.
• Model Coverage Audit: Ensure all target models are successfully collected each period; if a model is temporarily unavailable, replace it with an equivalent model and publicly explain.

8.2 Index Revision Policy

• Regular Revision: Review the indicator system, weights, and industry samples annually. Revisions will be publicly announced 30 days before the release of the first issue of the following year.
• Major Revision: If a fundamental change in AI large model technical architecture (e.g., RAG shifting to a new paradigm) may render indicators invalid, initiate an interim revision, announce it in advance, and provide explanations.

8.3 Independence Statement

The index compilation institution is independent of any enterprise or AI model provider in the evaluated industries, does not accept targeted funding, and ensures objective and fair evaluation results. Compilation funds come from the institution's own funds and public issuance revenue.

9. Interpretation and Usage Guide

9.1 Key Points for Index Interpretation

• CIII measures the "cognitive discourse power" of an industry in AI large models, which is not equivalent to the industry's actual technical strength or market share, but the long-term trends of the two should be positively correlated.
• The discourse power sub-item reflects the breadth and authority of industry knowledge presented by AI; the recommendation ranking sub-item reflects the AI's tendency toward that country's industry in decision support.
• Cross-industry comparisons should note differences in industry characteristics: consumer goods industries (e.g., new energy vehicles) have higher natural mention rates in C-end questions, while B2B industries like equipment manufacturing may excel in technical depth indicators.

9.2 Usage Scenarios

10. Limitations

• Model Representativeness Limitation: The index only reflects performance in tested models and cannot cover all AI models, especially vertical domain-specific models.
• Response Randomness: Generative AI outputs have inherent randomness; although multiple samplings reduce the impact, measurement errors still exist.
• Sensitivity to Technical Iteration: Large model version updates may cause short-term indicator fluctuations; judgments should be based on trends rather than single-quarter values.
• Language Barriers: Limited coverage of non-Chinese/English languages may affect a comprehensive assessment of Chinese industries' discourse power in some regional AI ecosystems.

11. Appendices

Appendix A: CIII Standardized Question Set (Examples)

Appendix B: CIII Authoritative Source Library (Excerpt)

Refer to GEO100:2026 Appendix B, with additions of industry standard-setting bodies (e.g., technical committees under ISO/IEC/ITU, China's National Standards Committee, IEEE SA, etc.), omitted.

Appendix C: Delphi Method Weight Determination Process

Detailed description of expert selection (over 5 each in industrial research, AI technology, and international communication fields), scoring rounds (three rounds), consistency test (Kendall's W ≥ 0.7), omitted.

Appendix D: Named Entity Recognition and Standard Citation Detection Rules

Provides NER model description and standard number regex matching rules, omitted.

Normative References

[1] United Nations Statistics Division. (2015). Handbook on Statistical Indicators.
[2] OECD. (2008). Handbook on Constructing Composite Indicators: Methodology and User Guide.
[3] ISO 10668:2010. Brand valuation – Requirements for monetary brand valuation.
[4] Pang Pei. (2025). Media-type GEO: A New Paradigm for Building Brand Trust Assets in the AI Era. Working Paper

[5] Pang Pei. (2026). Trust Verification, Cognitive Parasitism, and Dynamic Decay: Theoretical Extension and Model Deepening of AI Brand Equity (AIBE).
[6] National Advertising Research Institute. (2026). 2026 White Paper on AI Brand Equity Development.
[7] iResearch. (2026). Reshaping Search Visibility and Content Marketing in the AI Era.

Compiling Organizations

World Intelligence Organization (WIO), CNBNTV

Release Statement
This index is compiled in accordance with internationally accepted statistical standards. All rights reserved. Reproduction must cite the source. The index results do not constitute the sole basis for investment or policy decisions; users should make comprehensive judgments in conjunction with other information.