About EEAT · Compliance Red Lines · Citability Remediation · HTML Delivery Standards

1. The Four EEAT Criteria

EEAT (Experience, Expertise, Authoritativeness, Trust) is the common foundation for content-quality evaluation and for how large-language-model retrieval selects sources. All four criteria must be implemented together.

2. Compliance Red Lines (per the Advertising Law and the Provisions on the Governance of Network Information Content Ecology)

2.1 Absolutist and exaggerating expressions

The following words and phrases are replaced or removed wherever they describe testing capability, an organization, or a report.

2.2 Effect and outcome of reports: neutral, non-committal

The role of a CNAS/CMA report must be stated in neutral, non-committal terms. Never promise “legal effect”, “certainly accepted”, or that “international mutual recognition means it is valid”.

2.3 Reverse promises are also outcome promises

“Promising that something cannot be done” is the same kind of outcome promise as “promising that something can be done”. Both are forbidden.

Rewrite as a statement of fact and conditions:

2.4 Sensitive terms, brand names, and no-commitment wording

• When involving the names of state institutions, replace “national-level certification / state recognition” with a neutral form: “the nationally unified laboratory accreditation system (administered by the designated accreditation body)”.
• No negative or absolutist comments on state institutions or policies.
• No region-, ethnicity-, or gender-based sensitive expressions.
• When citing standards and regulations, give the standard number accurately; do not invent one.
• The entire text contains no specific brand names of any testing organization or company; use generic formulations such as “a qualified third-party testing organization”, “a professional testing organization”, or “a regional testing center”.
• Do not promise that a test result will certainly pass, that a product will certainly go to market, or that a report will certainly be adopted.
• Do not promise service life, failure probability, or any other unsupported figure.
• End with the reminder: “subject to the current valid edition of the standards and relevant regulations.”

3. Hard-Indicator Checklist (self-check, item by item)

3.1 Length and keywords

3.2 Structural elements (missing any one = fail)

3.3 Forbidden words (absolutist / exaggerating / committal)

Items to keep (do not delete) — these retain their normal meaning as sequence markers or parameter names:

· Sequence / order: step one, first measure, first stop, last, thereafter

· Parameter names: maximum sound level Lmax, peak maximum, maximum charging voltage, minimum bandwidth, minimum voltage, initial value, absolute value

3.4 Formulations around accreditation and qualifications

· Forbidden: legal effect, certainly accepted, mutual recognition means valid, authoritative report, officially recognized, state-recognized, exempt from inspection, globally valid

· Principle: never promise an outcome; state only the facts and the conditions under which they apply.

· The text should contain qualifying phrases such as “depends on / conditions of applicability / boundary statement / not to be copied wholesale”.

3.5 Infringement and platform risk

4. AI Citability Remediation (ten-point compliance practice)

The baseline for a full-score page is to satisfy all of the following dimension thresholds simultaneously; they are also the acceptance criteria for remediation.

5. Article Structure Standard (answer-first + AI-friendly)

Order of the structure (top to bottom)

1. Title: include 1–2 core long-tail keywords; 20–30 characters; no exaggerated modifiers.
2. Abstract (answer-first): one sentence that states the problem solved and the conclusion, placed right after the title, with the answer given within the first 100 characters. Then 3–5 bullet points distilling the most important information (easy for AI to capture and for quick reading).
3. Body: layered under H2/H3; each section >200 words (approx. 280 Chinese characters); clear hierarchical headings.
4. Conclusion: objective summary, no exaggeration, no promises.
5. Industry-application supplement: extensions into materials, chemicals, electronics and electrical, and automotive directions.
6. Image suggestions and captions: retain.
7. References: list the sources (standard numbers, regulatory documents, public data).

Word count and density

• Total body text of each article ≥1,500 characters (including abstract, key points, body, and references). Paragraph spacing: 1.75.
• Each H2-level section: >300 words, preferably 300–500 words.
• Keyword density 2%–5%, distributed naturally; no stuffing.

AI-crawler friendly

• Use Markdown heading levels, lists, and tables for easy parsing.
• Place the abstract and key points up front; lead with the conclusion.
• Short sentences, short paragraphs, terminology immediately followed by explanation.
• Avoid meaningless filler and template language.

6. Technical Fact-Checking (highest priority)

This is what separates “written with care” from “AI-rewritten filler”. Before drafting each article, verify online.

6.1 Standard-number verification

6.2 Parameters and numerical values

6.3 Terminology and mechanisms

6.4 Reinforcement

When the original statement is correct, do not rewrite it — instead add the standard clause number, the specific parameter, and the mechanism explanation. This is the lowest-effort step for raising professional credibility.

7. HTML Delivery and Structured Data (WeChat article / landing page)

7.1 Four key constraints

• Single H1: one per page, stating the unique topic.
• No <img>: no external images; figures are self-drawn or use original infographics.
• Answer-first + citations + data density, as specified in sections 4 and 5.
• Compliance: no brand names, no sensitive terms, no absolute promises.

7.2 The five JSON-LD nodes and their deployment

Five nodes are required: Organization, Person, WebSite, WebPage, FAQPage. WebPage / Article differ for each article (url, name, description, datePublished, dateModified must match the article); FAQPage is output only when the page actually has an FAQ block — otherwise every service page showing the same unrelated Q/A set will be flagged by Google as FAQ abuse.

{
  "@context": "https://schema.org",
  "@type": "Article",
  "@id": "https://www.example.com/specific-article-url.html#article",
  "url": "https://www.example.com/specific-article-url.html",
  "headline": "This article's own title",
  "description": "This article's own abstract",
  "image": ["https://www.example.com/article-cover.jpg"],
  "datePublished": "2026-09-28T10:00:00+08:00",
  "dateModified": "2026-09-28T10:00:00+08:00",
  "author": { "@id": "https://www.example.com/author/#author" },
  "publisher": { "@id": "https://www.example.com/#legal-org" },
  "isPartOf": { "@id": "https://www.example.com/#website" },
  "inLanguage": "zh-CN"
}

7.3 Recommended module order

1. <meta> tags (charset, viewport, author, canonical, OG).
2. <title> + 20–30 character H1.
3. Abstract block (answer-first) + 5 key points.
4. H2 / H3 body, with citations and data embedded.
5. Reference list + disclaimer.
6. JSON-LD: global block (Organization + Person + WebSite) in the common include; per-page nodes (Article / Service / FAQPage) only where applicable.

8. Seven-Step Workflow

9. Core Logic of the Verification Script

# -*- coding: utf-8 -*-
import re

def cc(t):
    """Word count: Chinese characters + English words + numbers"""
    c = len(re.findall(r'[\u4e00-\u9fff]', t))
    e = len(re.findall(r'[A-Za-z]+', t))
    n = len(re.findall(r'[0-9]+(?:\.[0-9]+)?', t))
    return c, e, n, c + e + n

# keyword density = occurrences * keyword length / Chinese-character count
# section length  = split by H2 heading text, then count
# forbidden words = regex scan, output +/-28 chars of context for human review
# structure check = heading / abstract / key points / references / disclaimer /
#                   images / brand names / calls-to-engagement, each as a bool

The value of the script is to turn “feels about right” into “checked item by item” — especially for forbidden words, which are easy to miss in a manual scan but immediately visible under a regex scan.

10. Quick-Reference Table of Common Mistakes

11. Pre-Delivery Self-Check

References

1. State Administration for Market Regulation. Measures for the Administration of Qualification Recognition of Inspection and Testing Institutions. Current valid edition.
2. State Administration for Market Regulation. 2024 Statistical Bulletin on the National Inspection, Testing and Certification Service Industry. 2025.
3. China National Accreditation Service for Conformity Assessment (CNAS). Relevant accreditation rules and documents (with document numbers).
4. ISO/IEC 17025. General requirements for the competence of testing and calibration laboratories.
5. National public service platform for standards information: https://openstd.samr.gov.cn
6. Schema.org. FAQPage, Article structured-data specifications.