seo

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

SEO

A full-spectrum search visibility skill. It treats traditional SEO, Answer Engine Optimization (AEO), and Generative Engine Optimization (GEO) as overlapping work across different search and answer surfaces, not as separate collections of ranking hacks.

一个覆盖全领域的搜索可见性 skill。它将传统 SEO、Answer Engine Optimization (AEO) 和 Generative Engine Optimization (GEO) 视为跨越不同搜索与答案场景的重叠工作，而不是一组组彼此独立的排名技巧。

Operating model

运营模型

Scope the surface and outcome. Name the provider, product or search surface, audience, entity, questions, business outcome, and exclusions. Define whether success means crawl access, index eligibility, retrieval, mention, citation, citation correctness, share of voice, referral, or conversion.
Inspect and research. Audit the public/rendered page and technical delivery. Read current first-party provider guidance and authoritative subject sources. Treat practitioner claims and vendor studies as hypotheses unless methods and scope support more.
Map intent to canonical content. Use topics and question clusters, but avoid manufacturing near-duplicate pages. Assign one canonical answer location, entity ownership, evidence, freshness owner, and internal links.
Improve people-first content and structure. Put a concise answer near the relevant heading, then supporting evidence, qualifications, and useful detail. Preserve natural prose, distinct point of view, accessibility, and human value.
Implement only supported controls. Fix crawlability, indexability, metadata, internal links, page experience, textual content, structured-data parity, sitemaps, freshness, and provider-specific crawler or preview controls. Optional files such as
```
llms.txt
```
or Markdown representations are provider-scoped proposals, not universal requirements.
Measure at the correct boundary. Freeze prompts and versions, capture exact answers and citations, score citation correctness, and use provider-native reports where available. Separate implementation evidence from observed visibility and causal claims.
Verify and learn. Recheck the rendered/public boundary, validate structured data, inspect search-console or provider evidence, and run bounded one-variable experiments. Do not declare success from HTTP 200, parseable JSON-LD, a single answer, or a third-party score.

界定场景与结果。 明确平台提供方（provider）、产品或搜索场景、受众、实体、问题、业务结果和排除项。定义成功究竟意味着抓取访问、索引资格、检索、提及、引用、引用正确性、声量份额（share of voice）、引荐流量还是转化。
检查与调研。 审计公开/渲染后页面和技术交付。阅读当前平台提供方的第一方指南和权威主题来源。除非方法和适用范围能够支持更强结论，否则将从业者主张和厂商研究视为假设。
将意图映射到 canonical 内容。 使用主题和问题簇，但避免制造近似重复页面。指定唯一的 canonical 答案位置、实体归属、证据、新鲜度负责人和内部链接。
改进以人为本的内容与结构。 在相关标题附近给出简明答案，随后提供支持性证据、限定条件和有用细节。保留自然文风、独特观点、无障碍性和人类价值。
仅实施有依据的控制措施。 修复可抓取性、可索引性、metadata、内部链接、页面体验、文本内容、结构化数据一致性、sitemaps、新鲜度，以及特定平台的 crawler 或预览控制。可选文件（如
```
llms.txt
```
或 Markdown 表示形式）属于面向特定平台的提案，而不是通用要求。
在正确边界进行衡量。 冻结 prompts 及其版本，捕获准确答案与引用，评估引用正确性，并在可用时使用平台原生报告。将实施证据与观察到的可见性及因果主张分开。
验证并学习。 重新检查渲染后/公开边界，验证结构化数据，检查 Search Console 或平台证据，并进行范围受控的一次一变量实验。不要仅凭 HTTP 200、可解析的 JSON-LD、单个答案或第三方评分就宣告成功。

Terminology and boundaries

术语与边界

SEO is the umbrella: improving a site's eligibility, discoverability, interpretation, and useful visibility in search systems.
AEO is a stakeholder label for answer-oriented work, including direct answers, snippets, knowledge surfaces, and answer-engine inclusion.
GEO is a stakeholder label for visibility in generated answers, especially being selected, cited, or factually absorbed into a synthesized response. The term originated in the 2024 KDD paper by Aggarwal et al.; it does not establish a universal algorithm.
LLMO and AI-search optimization are overlapping labels. Preserve the target organization's term, then define the measurable outcome and provider scope.
These terms do not guarantee ranking, inclusion, citation, traffic, recommendation, or conversion. A mention is not a citation, a citation is not proof of correctness, and a citation is not a click.

SEO 是总括概念：提升网站在搜索系统中的收录资格、可发现性、可理解性和有效可见性。
AEO 是利益相关方用于称呼答案导向工作的标签，包括直接答案、摘要片段、知识展示面和答案引擎收录。
GEO 是利益相关方用于称呼生成式答案可见性的标签，尤其指内容被选中、被引用，或内容中的事实被吸收到综合生成的响应中。该术语源自 Aggarwal 等人于 2024 年发表的 KDD 论文；它并不代表存在一种通用算法。
LLMO 和 AI-search optimization 是相互重叠的标签。保留目标组织使用的术语，然后定义可衡量的结果和平台范围。
这些术语并不保证排名、收录、引用、流量、推荐或转化。提及不等于引用，引用不能证明正确，引用也不等于点击。

Evidence rules

证据规则

Use these labels in plans and reports:

Primary documentation: provider, standards body, schema vocabulary, or tool owner describes its own behavior.
Observed: recorded crawl, rendered page, provider dashboard, exact answer, URL, or reproducible local result.
Independent study: disclosed method and dataset, with generalization limits stated.
Vendor-reported: useful for hypothesis generation, not a universal rule.
Inference: reasoned interpretation beyond direct observation.
Unresolved: provider-dependent, stale, contradicted, or not verified.

Reject claims such as “FAQ schema guarantees citations,” “short paragraphs are required by all engines,” “

llms.txt

is a search standard,” or “allowing a crawler guarantees inclusion.” Record source URL, access date, provider scope, exact support, what it does not prove, confidence, and refresh trigger for every material claim.

在计划和报告中使用以下标签：

Primary documentation（第一方文档）： 平台提供方、标准机构、schema 词汇表或工具所有者对自己行为的说明。
Observed（观察结果）： 已记录的抓取、渲染后页面、平台控制台、准确答案、URL 或可复现的本地结果。
Independent study（独立研究）： 具有已披露的方法和数据集，并说明泛化限制。
Vendor-reported（厂商报告）： 有助于形成假设，但不是通用规则。
Inference（推断）： 基于证据、超出直接观察范围的合理解读。
Unresolved（未解决）： 取决于平台、已过时、相互矛盾或尚未验证。

拒绝诸如“FAQ schema 保证获得引用”“所有引擎都要求使用短段落”“

llms.txt

是搜索标准”或“允许某个 crawler 就保证收录”等主张。对于每项实质性主张，记录来源 URL、访问日期、平台范围、确切支持内容、其不能证明的事项、置信度及刷新触发条件。

Reference routing

参考资料路由

Need	Read
Crawlability, indexability, robots, sitemaps, performance, canonicals, mobile, HTTPS	`references/technical-seo.md`
Titles, descriptions, headings, content quality, links, images	`references/onpage-seo.md`
Schema.org, JSON-LD, rich-result eligibility, visible parity	`references/schema-markup.md` and `references/structured-data.md`
Topics, question clusters, entities, answer blocks, evidence architecture	`references/content-and-entity-architecture.md`
Full answer/generative implementation sequence and completion gate	`references/implementation-playbook.md`
Provider guidance, crawler identities, robots and preview controls	`references/platform-guidance.md` and `references/discovery-and-freshness.md`
`llms.txt` , Markdown delivery, content negotiation, provider support	`references/agent-readable-content.md`
Outcome definitions, confidence, and rejected claims	`references/evidence-boundaries.md`
Prompt sets, citation logs, metrics, experiments, and confounders	`references/measurement-and-experimentation.md`
Ghost metadata and injection	`references/ghost-metadata.md`
Content strategy, topic clusters, keywords, gaps, SERP features	`references/content-strategy-seo.md`
Source URLs, access dates, authority tiers, and refresh notes	`references/source-index.md`

需求	阅读
可抓取性、可索引性、robots、sitemaps、性能、canonical、移动端、HTTPS	`references/technical-seo.md`
标题、描述、标题层级、内容质量、链接、图片	`references/onpage-seo.md`
Schema.org、JSON-LD、rich result 资格、与可见内容一致	`references/schema-markup.md` 和 `references/structured-data.md`
主题、问题簇、实体、答案块、证据架构	`references/content-and-entity-architecture.md`
完整的答案/生成式实施顺序和完成门槛	`references/implementation-playbook.md`
平台指南、crawler 身份、robots 和预览控制	`references/platform-guidance.md` 和 `references/discovery-and-freshness.md`
`llms.txt` 、Markdown 交付、content negotiation、平台支持	`references/agent-readable-content.md`
结果定义、置信度和被拒绝的主张	`references/evidence-boundaries.md`
prompt 集、引用日志、指标、实验和混杂因素	`references/measurement-and-experimentation.md`
ghost metadata 和注入	`references/ghost-metadata.md`
内容策略、主题集群、关键词、内容缺口、SERP features	`references/content-strategy-seo.md`
来源 URL、访问日期、权威层级和刷新说明	`references/source-index.md`

Scripts and templates

脚本和模板

Run from the skill directory:

bash

python3 scripts/aeo_audit.py <page.html-or-URL> --json
python3 scripts/build_prompt_matrix.py <topics.json> --output prompt-matrix.json
python3 -m pytest scripts/test_aeo_scripts.py

The scripts are read-only and use the Python standard library. They inspect source HTML; they do not execute JavaScript, call an LLM, publish, submit URLs, or change crawler policy. Templates cover implementation plans, question clusters, citation observations, optional

llms.txt

, and crawler-policy decisions.

在 skill 目录中运行：

bash

python3 scripts/aeo_audit.py <page.html-or-URL> --json
python3 scripts/build_prompt_matrix.py <topics.json> --output prompt-matrix.json
python3 -m pytest scripts/test_aeo_scripts.py

这些脚本是只读的，并使用 Python 标准库。它们会检查源 HTML；不会执行 JavaScript、调用 LLM、发布内容、提交 URL，也不会更改 crawler 策略。模板涵盖实施计划、问题簇、引用观察记录、可选的

llms.txt

和 crawler 策略决策。

Audit output

审计输出

Use

assets/audit-report-template.md

and distinguish:

Observed findings: what the inspected page, response, dashboard, or answer actually shows.
Recommended changes: proposed actions with owner, risk, expected mechanism, and verification.
Provider scope: which engine or search surface the evidence applies to.
Status: implemented, verified, observed, inferred, or unresolved.

使用

assets/audit-report-template.md

，并区分：

Observed findings（观察到的发现）： 受检页面、响应、控制台或答案实际显示的内容。
Recommended changes（建议变更）： 包含负责人、风险、预期机制和验证方式的拟议操作。
Provider scope（平台范围）： 证据适用的 engine 或搜索场景。
Status（状态）： 已实施、已验证、已观察、推断或未解决。

When not to use

不适用的场景

Do not use this skill alone for only mechanical copy-editing, ordinary article writing, CMS administration, or a generic request to “rank better” without a defined target, evidence boundary, or measurable outcome. Route those to the relevant writing, copy-editing, CMS, or product skill.

不要单独使用此 skill 来完成纯机械式文案编辑、普通文章写作、CMS 管理，或在缺少明确目标、证据边界或可衡量结果时仅提出“排名更好”的要求。将这些请求转交给相应的写作、文案编辑、CMS 或产品 skill。

Portability

可移植性

Use the host agent's normal mechanisms to load references, templates, and scripts. Do not assume a particular profile system, orchestrator, memory service, CMS, search console, or provider API.

使用宿主 Agent 的常规机制加载 references、templates 和 scripts。不要假定存在特定的 profile system、orchestrator、memory service、CMS、Search Console 或 provider API。