keyword-research

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

When this skill is activated, always start your first response with the 🧢 emoji.

激活本Skill后，首次回复请务必以🧢表情开头。

Keyword Research

关键词研究

Quick start (5 steps): Seed keywords > Expand with tools/autocomplete > Classify intent for each keyword > Tri-score (organic + AEO + GEO) > Cluster by topic with surface annotations. See the common tasks below for details.

Keyword research is the foundation of all organic and AI search strategy. It is the process of discovering what words and phrases people type into search engines and AI assistants, understanding why they search (intent), and evaluating which of three surfaces - organic results, answer engine features, or AI-generated citations - offers the best opportunity for each keyword.

In 2026, keyword research must account for three surfaces simultaneously:

Organic blue links (SEO) - traditional rankings on Google, Bing, etc.
Answer engine features (AEO) - featured snippets, People Also Ask, voice results
AI-generated citations (GEO) - Google AI Overviews, ChatGPT Search, Perplexity

This skill covers the full research workflow - from seed topic to prioritized, tri-surface-scored keyword report. It tells you WHAT to target and WHERE the opportunity is. For HOW to optimize content once you have chosen your targets, use the companion skills

aeo-optimization

(snippet/PAA formatting) and

geo-optimization

(AI citation optimization).

快速入门（5步）： 种子关键词 > 借助工具/自动补全拓展关键词 > 为每个关键词分类意图 > 三渠道评分（自然搜索+AEO+GEO） > 按主题聚类并标注渠道。详情请见下方常见任务。

关键词研究是所有自然搜索和AI搜索策略的基础。它是挖掘人们在搜索引擎和AI助手输入的词汇与短语、理解其搜索原因（意图），并评估自然搜索结果、问答引擎功能或AI生成引用这三个渠道中，哪个渠道对每个关键词的机会最大的过程。

到2026年，关键词研究必须同时考虑三个渠道：

自然搜索蓝色链接（SEO） - Google、Bing等平台的传统排名
问答引擎功能（AEO） - 精选摘要、People Also Ask（PAA）、语音搜索结果
AI生成引用（GEO） - Google AI Overview、ChatGPT Search、Perplexity

本Skill覆盖完整的研究流程——从种子主题到经过优先级排序、三渠道评分的关键词报告。它会告诉你要瞄准什么目标，以及机会在哪里。一旦确定目标后，如需了解如何优化内容，请使用配套Skill

aeo-optimization

（摘要/PAA格式优化）和

geo-optimization

（AI引用优化）。

When to use this skill

何时使用本Skill

Trigger this skill when the user:

Wants to find keywords for a new website, product page, or blog
Asks to analyze search intent for a keyword list
Needs to group keywords into topic clusters or content pillars
Wants to discover competitor keyword gaps or ranking opportunities
Asks to find long-tail variations of a seed keyword
Needs to prioritize a list of keywords by opportunity or difficulty
Wants to understand what SERP features appear for a target keyword
Asks to detect keyword cannibalization across existing pages
Wants to evaluate keywords for snippet opportunity or featured snippet potential
Asks to assess AI Overview presence for target keywords
Needs a tri-surface keyword report scoring organic, AEO, and GEO opportunity
Wants to understand which keywords AI search engines answer vs. defer

Do NOT trigger this skill for:

Paid search (PPC/Google Ads) bid strategy - ad-specific match types, Quality Scores, and CPC optimization are a different domain
Brand naming or tagline development - that is copywriting, not search research
Formatting content to win snippets or PAA - that is
```
aeo-optimization
```
Making content more citable by AI engines - that is
```
geo-optimization
```

当用户有以下需求时，触发本Skill：

为新网站、产品页或博客寻找关键词
要求分析关键词列表的搜索意图
需要将关键词分组为主题集群或内容支柱
想要挖掘竞品关键词缺口或排名机会
要求寻找种子关键词的长尾变体
需要按机会或难度对关键词列表进行优先级排序
想要了解目标关键词对应的SERP功能有哪些
要求检测现有页面间的关键词 cannibalization
想要评估关键词的摘要获取机会或精选摘要潜力
要求评估目标关键词的AI Overview展示情况
需要一份包含自然搜索、AEO和GEO机会评分的三渠道关键词报告
想要了解哪些关键词会被AI搜索引擎回答，哪些会被跳过

以下情况请勿触发本Skill：

付费搜索（PPC/Google Ads）出价策略——广告特定匹配类型、质量得分和CPC优化属于不同领域
品牌命名或标语开发——这属于文案创作，而非搜索研究
优化内容格式以获取摘要或PAA——这属于
```
aeo-optimization
```
的范畴
优化内容以提升被AI引擎引用的概率——这属于
```
geo-optimization
```
的范畴

Key principles

核心原则

Search intent is more important than volume - A keyword with 500 monthly searches and clear transactional intent will drive more revenue than a 50,000-search keyword that is purely informational. Always qualify intent before qualifying volume.
Cluster keywords by topic, not individual pages - One page should own a cluster of semantically related terms. Building one page per keyword creates duplication, splits authority, and fragments the user experience.
The SERP + AI Overview is the source of truth - No tool tells you more about what Google wants to rank than the current top 10 results and whether an AI Overview fires. Content type, length, format, featured snippet presence, and AI Overview citations all reveal the implicit standard for a keyword.
Long-tail keywords convert better - Longer, more specific queries have lower volume but higher purchase intent and lower competition. A content strategy built on long-tail clusters outperforms chasing high-volume head terms in most niches.
Competitor gaps reveal the fastest wins - Finding keywords where competitors rank in positions 4-15 (or not at all) is faster than trying to beat them on keywords where they dominate. Gaps are the entry points.
Every keyword has three surfaces to evaluate - A keyword that looks mediocre for organic ranking may have excellent snippet opportunity (AEO) or strong AI citation potential (GEO). Evaluating all three surfaces prevents blind spots and reveals non-obvious wins that single-surface research misses.
Research and optimization are separate phases - This skill identifies WHAT to target and WHERE the opportunity is. Optimization (HOW to format for snippets, HOW to boost AI citations) is a downstream activity handled by
```
aeo-optimization
```
and
```
geo-optimization
```
. Do not mix phases - complete research before starting optimization.

搜索意图比搜索量更重要 - 一个月搜索量500次但意图明确的交易型关键词，能带来的收入远高于月搜索量50000次但仅为信息型的关键词。在评估搜索量之前，务必先确认意图。
按主题聚类关键词，而非为每个关键词单独建页 - 一个页面应覆盖一组语义相关的词汇。为每个关键词单独建页会导致内容重复、分散权重，并割裂用户体验。
SERP + AI Overview是判断标准 - 没有任何工具比当前排名前10的结果以及是否展示AI Overview更能告诉你Google想要的排名内容。内容类型、长度、格式、精选摘要存在情况和AI Overview引用来源，都揭示了关键词的隐性标准。
长尾关键词转化率更高 - 更长、更具体的查询搜索量较低，但购买意图更强，竞争也更小。在大多数细分领域，基于长尾关键词集群构建的内容策略，表现优于追逐高搜索量的核心关键词。
竞品缺口是快速取胜的关键 - 找到竞品排名在4-15位（或未排名）的关键词，比试图在他们占据主导的关键词上击败他们更快。缺口是切入点。
每个关键词都要评估三个渠道 - 一个在自然搜索排名中表现平平的关键词，可能在摘要获取机会（AEO）或AI引用潜力（GEO）上表现出色。评估所有三个渠道可以避免盲区，发现单渠道研究错过的非显性机会。
研究与优化是独立的阶段 - 本Skill确定要瞄准的目标和机会所在。优化（如何为摘要格式化内容、如何提升AI引用）是后续环节，由
```
aeo-optimization
```
和
```
geo-optimization
```
处理。不要混淆阶段——完成研究后再开始优化。

Core concepts

核心概念

Search intent taxonomy classifies every keyword into one of four categories based on what the searcher is trying to accomplish. Informational intent ("how does X work", "what is Y") signals content and education needs. Navigational intent ("brand name", "site login") signals the user knows where they want to go. Transactional intent ("buy X online", "X pricing", "X discount code") signals readiness to act. Commercial investigation ("best X", "X vs Y", "X review") sits between informational and transactional - the user is evaluating options before deciding. See

references/search-intent-mapping.md

for detailed classification guidance including intent-to-surface mapping.

Keyword difficulty (KD) is a 0-100 score estimating how hard it is to rank on page one for a keyword, based primarily on the backlink authority of the current top-ranking pages. High difficulty does not mean impossible - it means you need more authority, better content, or a more specific angle to win. Treat KD as a relative filter, not an absolute gate.

Search volume vs. traffic potential are related but different. Search volume is the average monthly searches for one keyword. Traffic potential is the estimated traffic the top-ranking page receives for the entire cluster of keywords it ranks for. A keyword with 200 monthly searches may have traffic potential of 2,000 if the ranking page captures dozens of related terms. Always evaluate traffic potential over raw volume.

Keyword cannibalization occurs when two or more pages on the same site compete for the same keyword, splitting ranking signals and confusing Google about which page to surface. Symptoms include ranking oscillation, positions that drop when publishing new content, and two pages from the same domain appearing for the same query. Resolve by merging, redirecting, or clearly differentiating the pages.

Tri-surface keyword scoring evaluates every keyword across three surfaces: organic opportunity (0-10), AEO opportunity (0-10), and GEO opportunity (0-10). The composite score (0-30) reveals total opportunity, and the priority surface tells you where to focus optimization efforts. See

references/tri-surface-scoring.md

for the full scoring rubrics and worked examples.

SERP feature landscape - SERP features (featured snippets, PAA, video carousels, image packs, shopping results, AI Overviews) are research signals, not just ranking features. Their presence or absence tells you which surfaces are active for a keyword. A keyword with a featured snippet and an AI Overview has tri-surface potential. A keyword with only shopping results is organic-only. Map features during research to inform scoring.

AI search query patterns - AI search engines (Google AI Overviews, ChatGPT Search, Perplexity) do not answer every query. They tend to fire on informational, comparison, and multi-step queries while skipping navigational, transactional, and simple factual queries. Understanding these patterns helps predict GEO opportunity during research. See

references/search-intent-mapping.md

for the full AI Overview intent pattern table.

搜索意图分类体系 根据搜索者的目标，将每个关键词分为四类。信息型意图（如“X如何工作”、“Y是什么”）表示用户有内容和学习需求。导航型意图（如“品牌名称”、“网站登录”）表示用户知道自己想去哪里。交易型意图（如“在线购买X”、“X价格”、“X折扣码”）表示用户准备采取行动。商业调研型意图（如“最佳X”、“X vs Y”、“X评测”）介于信息型和交易型之间——用户正在评估选项，之后才会做决定。如需详细分类指南，包括意图与渠道的对应关系，请查看

references/search-intent-mapping.md

。

关键词难度（KD） 是一个0-100的分数，主要根据当前排名靠前页面的外链权重，估算进入关键词首页排名的难度。高难度并不意味着不可能——它意味着你需要更高的权重、更好的内容或更独特的角度才能取胜。将KD视为相对筛选标准，而非绝对门槛。

搜索量与流量潜力 相关但不同。搜索量是单个关键词的平均月搜索次数。流量潜力是排名靠前的页面因整个关键词集群获得的预估流量。一个月搜索量200次的关键词，对应的排名页面可能覆盖数十个相关词汇，带来2000次月访问量。始终评估流量潜力，而非原始搜索量。

关键词 cannibalization 指同一网站的两个或多个页面针对同一关键词竞争，这会分散排名信号，让Google无法确定应该展示哪个页面。症状包括排名波动、发布新内容后排名下降、同一域名的两个页面出现在同一查询结果中。解决方法是合并页面、设置重定向，或明确区分页面的定位。

三渠道关键词评分 从三个渠道评估每个关键词：自然搜索机会（0-10）、AEO机会（0-10）和GEO机会（0-10）。综合得分（0-30）显示总机会，优先级渠道告诉你优化工作应聚焦何处。完整评分标准和示例请见

references/tri-surface-scoring.md

。

SERP功能格局 - SERP功能（精选摘要、PAA、视频轮播、图片包、购物结果、AI Overview）不仅是排名功能，也是研究信号。它们的存在或缺失告诉你关键词对应的活跃渠道。同时拥有精选摘要和AI Overview的关键词具有三渠道潜力。仅显示购物结果的关键词属于纯自然搜索渠道。研究期间需梳理这些功能，为评分提供依据。

AI搜索查询模式 - AI搜索引擎（Google AI Overview、ChatGPT Search、Perplexity）不会回答所有查询。它们通常会响应对信息型、对比型和多步骤查询，而跳过导航型、交易型和简单事实型查询。了解这些模式有助于在研究期间预测GEO机会。完整的AI Overview意图模式表请见

references/search-intent-mapping.md

。

Common tasks

常见任务

1. Map search intent for a keyword list

1. 为关键词列表映射搜索意图

For each keyword, classify it using the four-type taxonomy. Apply this decision order:

Check modifiers first - Words like "buy", "order", "coupon", "discount" signal transactional. Words like "best", "top", "review", "vs", "alternative" signal commercial investigation. Words like "how", "what", "why", "guide", "tutorial" signal informational. Brand name only = navigational.
When modifiers are absent, check the SERP - Look at the top 3 results. Are they product pages, comparison articles, definitions, or brand homepages? The content type Google rewards reveals the intent.
Assign a primary intent and note a secondary if relevant - Many keywords blend types. "Best project management software" is primarily commercial investigation with transactional secondary (the user may click through to pricing).

Output format: a table with columns

keyword | intent | confidence | content_type | snippet_type | ai_overview_present

The last two columns feed into AEO and GEO scoring. Record snippet type as paragraph/list/table/none and AI Overview presence as yes/sometimes/no.

See

references/search-intent-mapping.md

for the full classification guide.

使用四类分类体系为每个关键词分类。遵循以下决策顺序：

先检查修饰词 - 诸如“购买”、“订购”、“优惠券”、“折扣”等词表示交易型意图。诸如“最佳”、“顶级”、“评测”、“vs”、“替代方案”等词表示商业调研型意图。诸如“如何”、“什么”、“为什么”、“指南”、“教程”等词表示信息型意图。仅品牌名称属于导航型意图。
若无修饰词，查看SERP - 查看排名前3的结果。它们是产品页、对比文章、定义页还是品牌首页？Google推崇的内容类型揭示了意图。
分配主要意图，如有相关则标注次要意图 - 许多关键词混合了多种类型。“最佳项目管理软件”主要是商业调研型意图，附带交易型次要意图（用户可能点击查看价格）。

输出格式：包含以下列的表格

keyword | intent | confidence | content_type | snippet_type | ai_overview_present

最后两列用于AEO和GEO评分。记录摘要类型为段落/列表/表格/无，AI Overview展示情况为是/有时/否。

完整分类指南请见

references/search-intent-mapping.md

。

2. Build a keyword cluster from a seed topic

2. 从种子主题构建关键词集群

Start with one seed keyword and expand outward:

Generate variants - Use a keyword tool to pull: questions (People Also Ask), autocomplete suggestions, related searches, and lexical variants. For the seed "project management software", variants include "best project management tools", "project management app for teams", "free project management software", etc.
Group by SERP overlap - Keywords that return the same top-ranking URLs belong in the same cluster. If "project management software" and "task management tool" return 6 of the same top-10 results, one page can rank for both.
Identify the primary keyword - The one with the highest traffic potential becomes the primary term (used in title, H1, URL). All others are secondary terms woven into subheadings and body copy.
Name the cluster - Give it a descriptive label: "project management software - top-of-funnel commercial". This label drives content brief decisions.
Annotate with dominant surface - Score the cluster's keywords using tri-surface scoring and assign a surface label: [ORG], [AEO], [GEO], or [AEO+GEO]. This tells the content team which optimization approach to take after writing.

See

references/keyword-clustering.md

for semantic, SERP-based, modifier-based, and surface-aware clustering methods.

从一个种子关键词开始，向外拓展：

生成变体 - 使用关键词工具获取：问题（People Also Ask）、自动补全建议、相关搜索和词汇变体。对于种子关键词“项目管理软件”，变体包括“最佳项目管理工具”、“团队用项目管理应用”、“免费项目管理软件”等。
按SERP重叠分组 - 返回相同排名靠前URL的关键词应归为同一集群。如果“项目管理软件”和“任务管理工具”的前10个结果中有6个相同，那么一个页面即可同时为这两个关键词排名。
确定主关键词 - 流量潜力最高的关键词成为主关键词（用于标题、H1、URL）。其他均为次要关键词，可融入副标题和正文中。
命名集群 - 为集群起一个描述性标签：“项目管理软件 - 漏斗顶部商业调研型”。该标签将指导内容brief的制定。
标注主导渠道 - 使用三渠道评分对集群中的关键词打分，分配渠道标签：[ORG]、[AEO]、[GEO]或[AEO+GEO]。这会告诉内容团队写完内容后应采用哪种优化方法。

语义聚类、基于SERP的聚类、基于修饰词的聚类和渠道感知聚类方法请见

references/keyword-clustering.md

。

3. Tri-surface scoring

3. 三渠道评分

This is the signature task of modern keyword research. For each keyword (or cluster), produce scores across all three surfaces.

Process:

Gather raw data - For each keyword, collect: search volume, traffic potential, KD, SERP features present, featured snippet format and holder, PAA count, and AI Overview presence (check Google, optionally ChatGPT Search and Perplexity). See
```
references/tool-specific-workflows.md
```
for tool-by-tool data collection steps.
Score organic (0-10) - Based on traffic potential, KD inversion, intent-business alignment, and content gap existence.
Score AEO (0-10) - Based on snippet presence, format match, PAA count, voice search likelihood, and current holder strength. Score 0 for navigational and most transactional keywords where snippets don't fire.
Score GEO (0-10) - Based on AI Overview trigger, citation density, query type match, entity relevance, and content uniqueness. Score 0 when no AI Overview fires.
Calculate composite - Sum the three scores (0-30). Apply business-goal weighting if needed.
Assign priority surface - Highest-scoring surface becomes the priority. If two are within 2 points and both above 5, mark as dual-surface opportunity.

See

references/tri-surface-scoring.md

for the complete rubrics, weighting tables, and three worked examples.

Output format:

keyword | intent | organic_score | aeo_score | geo_score | composite | priority_surface

这是现代关键词研究的标志性任务。为每个关键词（或集群）在三个渠道上打分。

流程：

收集原始数据 - 为每个关键词收集：搜索量、流量潜力、KD、存在的SERP功能、精选摘要格式和持有者、PAA数量，以及AI Overview展示情况（检查Google，可选ChatGPT Search和Perplexity）。各工具的数据收集步骤请见
```
references/tool-specific-workflows.md
```
。
自然搜索评分（0-10） - 基于流量潜力、KD反向指标、意图与业务的匹配度以及内容缺口的存在情况。
AEO评分（0-10） - 基于摘要存在情况、格式匹配度、PAA数量、语音搜索可能性和当前持有者的实力。导航型和大多数交易型关键词不会触发摘要，评分为0。
GEO评分（0-10） - 基于AI Overview触发情况、引用密度、查询类型匹配度、实体相关性和内容独特性。如果未触发AI Overview，评分为0。
计算综合得分 - 将三个分数相加（0-30）。如有需要，可根据业务目标加权。
分配优先级渠道 - 得分最高的渠道为优先级渠道。如果两个渠道得分相差在2分以内且均高于5分，则标记为双渠道机会。

完整评分标准、加权表和三个示例请见

references/tri-surface-scoring.md

。

输出格式：

keyword | intent | organic_score | aeo_score | geo_score | composite | priority_surface

4. Identify competitor keyword gaps (surface-aware)

4. 识别竞品关键词缺口（渠道感知型）

A keyword gap is a keyword where a competitor ranks in the top 20 but you do not.

Framework for surface-aware gap analysis:

Pull keyword rankings for 3-5 competitors using a tool (Ahrefs, Semrush). Export keywords where competitor is in positions 1-20.
Filter out keywords where your site already ranks positions 1-5 (already winning).
Filter for keywords matching your target intent.
Sort by traffic potential descending.
New: Surface gap analysis - For the top 50 gap keywords, check whether they trigger featured snippets and AI Overviews. A keyword gap where the competitor ranks organically but doesn't hold the snippet or isn't cited by AI is a multi-surface gap - you can potentially win the snippet or AI citation even if the organic position takes time to capture.
Cross-reference with your content inventory. Existing pages that can be optimized are quick wins; missing pages are content creation opportunities.

See

references/tool-specific-workflows.md

for tool-specific gap analysis steps.

关键词缺口指竞品排名在前20位，但你的网站未排名的关键词。

渠道感知型缺口分析框架：

使用工具（Ahrefs、Semrush）获取3-5个竞品的关键词排名。导出竞品排名1-20位的关键词。
过滤掉你的网站已排名1-5位的关键词（已获胜的关键词）。
筛选符合目标意图的关键词。
按流量潜力降序排序。
新增：渠道缺口分析 - 对前50个缺口关键词，检查它们是否触发精选摘要和AI Overview。如果竞品仅在自然搜索排名，但未持有摘要或未被AI引用，这就是多渠道缺口——即使自然搜索排名需要时间提升，你也有可能获取摘要或AI引用。
与你的内容库存交叉对比。可优化的现有页面是快速取胜的机会；缺失的页面是内容创作机会。

各工具的缺口分析步骤请见

references/tool-specific-workflows.md

。

5. Find long-tail variations with surface annotations

5. 寻找带渠道标注的长尾关键词变体

Long-tail keywords (typically 3+ words, lower volume, higher specificity) are easier to rank for and often signal stronger intent. To find them:

Question modifiers: "how to", "what is", "why does", "when should"
Qualifier modifiers: "for small business", "for beginners", "without X", "with Y"
Comparison modifiers: "vs", "alternative to", "better than", "instead of"
Location modifiers: city, region, "near me", "in [country]"
Feature modifiers: "free", "open source", "enterprise", "API", "integration"

Surface annotations for long-tail keywords:

Question-format long-tails ("how to X for Y") score high on AEO (snippet targets)
Comparison long-tails ("X vs Y for Z") score high on GEO (AI engines love comparisons)
Feature/qualifier long-tails ("X with API for enterprise") are usually organic-only
Location long-tails are almost always organic-only (local pack, not snippets or AI)

Target long-tail keywords with dedicated FAQ sections, comparison pages, or use-case landing pages. Annotate each with its likely priority surface.

长尾关键词（通常3个词以上，搜索量较低，针对性更强）更容易排名，且通常表示更强的意图。寻找长尾关键词的方法：

疑问修饰词：“如何”、“是什么”、“为什么”、“何时应该”
限定修饰词：“针对小企业”、“针对初学者”、“无需X”、“带有Y”
对比修饰词：“vs”、“替代方案”、“比更好”、“代替”
位置修饰词：城市、地区、“附近”、“在[国家]”
功能修饰词：“免费”、“开源”、“企业级”、“API”、“集成”

长尾关键词的渠道标注：

疑问格式的长尾关键词（如“如何为Y做X”）在AEO渠道得分高（摘要目标）
对比格式的长尾关键词（如“X vs Y for Z”）在GEO渠道得分高（AI引擎喜欢对比内容）
功能/限定修饰词的长尾关键词（如“带有API的企业级X”）通常仅适用于自然搜索渠道
位置修饰词的长尾关键词几乎仅适用于自然搜索渠道（本地包，而非摘要或AI）

为长尾关键词设置专门的FAQ板块、对比页面或用例落地页。为每个关键词标注其可能的优先级渠道。

6. Produce a keyword research report

6. 制作关键词研究报告

The keyword research report is the primary deliverable of this skill. It synthesizes all research into an actionable document.

Report structure:

Executive summary (1 paragraph) - Total keywords analyzed, top clusters, dominant surface opportunity, and 3 biggest findings.
Intent distribution - Pie chart or table showing the breakdown of informational, commercial, transactional, and navigational keywords in the dataset.
Surface opportunity map - Table showing how many keywords have their priority surface as organic, AEO, GEO, dual, or tri. This reveals whether the overall strategy should lean toward traditional SEO, snippet optimization, or AI citation.
Top keyword clusters - For each cluster: name, primary keyword, keyword count, average composite score, dominant surface, and recommended content type.
Quick wins - Keywords where you have the best ratio of opportunity to effort: striking-distance organic keywords (positions 5-15), snippet-eligible keywords with no current holder, and AI Overview queries citing weak sources you can displace.
Content recommendations - Map clusters to content calendar items: new pages, existing pages to optimize, FAQ additions, and comparison articles to create.
Full keyword data - Appendix table with all keywords and their scores. Use the spreadsheet template from
```
references/tool-specific-workflows.md
```
.

关键词研究报告是本Skill的主要交付成果。它将所有研究内容整合为一份可执行的文档。

报告结构：

执行摘要（1段）- 分析的关键词总数、顶级集群、主导渠道机会以及3个最重要的发现。
意图分布 - 饼图或表格展示数据集中信息型、商业调研型、交易型和导航型关键词的占比。
渠道机会地图 - 表格显示有多少关键词的优先级渠道为自然搜索、AEO、GEO、双渠道或三渠道。这揭示了整体策略应倾向于传统SEO、摘要优化还是AI引用优化。
顶级关键词集群 - 每个集群包含：名称、主关键词、关键词数量、平均综合得分、主导渠道和推荐内容类型。
快速取胜机会 - 机会与投入比最高的关键词：接近首页的自然搜索关键词（排名5-15位）、无当前持有者的摘要 eligible关键词、AI Overview引用弱来源可被你取代的查询。
内容建议 - 将集群映射到内容日历项：新页面、需优化的现有页面、需添加的FAQ、需创建的对比文章。
完整关键词数据 - 附录表格包含所有关键词及其得分。可使用
```
references/tool-specific-workflows.md
```
中的电子表格模板。

7. Detect keyword cannibalization

7. 检测关键词 cannibalization

Run a site search for the target keyword (

site:yourdomain.com "keyword phrase"

) and audit Google Search Console for pages sharing the same top query.

Diagnosis:

Two pages ranking for the same query: Check which page Google prefers (higher avg. position). The preferred page keeps the keyword; the other page is re-optimized for a different term or redirected.
Rankings oscillating week to week: Classic cannibalization signal. Consolidate the weaker page's content into the stronger one via a 301 redirect.
New page tanked the ranking of an existing page: Re-differentiate the new page's focus term or merge it back into the original.
Snippet cannibalization: If two of your pages alternate holding the featured snippet for the same query, Google may drop both. Consolidate to ensure one definitive page owns the snippet-eligible answer.

针对目标关键词进行站内搜索（

site:yourdomain.com "keyword phrase"

），并审核Google Search Console中共享同一顶级查询的页面。

诊断方法：

两个页面针对同一关键词排名：查看Google偏好哪个页面（平均排名更高）。保留偏好页面的关键词，重新优化另一个页面以针对其他关键词，或设置重定向。
排名每周波动：典型的cannibalization信号。通过301重定向将较弱页面的内容合并到较强页面中。
新页面导致现有页面排名下降：重新明确新页面的核心关键词，或将其合并回原页面。
摘要 cannibalization：如果你的两个页面交替持有同一查询的精选摘要，Google可能会同时放弃这两个页面。合并页面，确保一个明确的页面拥有摘要 eligible的答案。

Anti-patterns

反模式

Mistake	Why it's wrong	What to do instead
Chasing volume over intent	A high-volume keyword that doesn't match your buyer's stage sends irrelevant traffic that bounces	Filter by intent first, then sort by volume within the right intent category
One page per keyword	Creates thin, near-duplicate pages that split link equity and rarely rank	Cluster semantically related keywords to one page; build depth
Ignoring the SERP	Targeting a keyword without checking what type of content currently ranks leads to mismatched format	Always check the top 10 before writing a brief; match dominant content type
Targeting KD 70+ with a new site	New domains lack the authority to rank on competitive terms	Start with KD < 30 to earn rankings, traffic, and links; build up to harder terms
Skipping competitor gap analysis	Building content only from brainstorming misses proven opportunities	Always run a gap report before finalizing your content calendar
Never updating keyword research	Search behavior evolves; queries from 2 years ago may have shifted in intent or volume	Audit top content annually; refresh keyword targets based on current SERP data
Ignoring snippet and AI Overview presence	Treating all keywords as organic-only misses AEO and GEO opportunities that may be easier to win than organic rankings	Record SERP features and AI Overview status for every keyword during research
Treating every keyword as organic-only	Defaulting to traditional SEO without checking if a keyword is better won through snippets or AI citations	Run tri-surface scoring for all priority keywords; assign a priority surface
Scoring GEO when no AI Overview fires	Assigning GEO opportunity to a keyword where AI engines don't generate an answer wastes effort	Always verify AI Overview presence manually; GEO = 0 if no AI answer fires
Mixing research and optimization	Trying to format content for snippets or AI citations before finishing keyword research leads to premature decisions	Complete the full research workflow (seed > expand > classify > score > cluster) before any optimization work

错误做法	错误原因	正确做法
追逐搜索量而非意图	与买家阶段不匹配的高搜索量关键词会带来无关流量，导致高跳出率	先按意图筛选，再在符合意图的类别中按搜索量排序
一个页面对应一个关键词	导致内容单薄、近乎重复，分散链接权重，且难以排名	将语义相关的关键词聚类到一个页面；打造深度内容
忽略SERP	不查看当前排名内容类型就瞄准关键词，会导致内容格式不匹配	撰写brief前务必查看前10个结果；匹配主流内容类型
新网站瞄准KD 70+的关键词	新域名缺乏权重，无法在竞争激烈的关键词中排名	从KD < 30的关键词开始，积累排名、流量和链接；逐步挑战难度更高的关键词
跳过竞品缺口分析	仅靠头脑风暴创作内容会错过已验证的机会	最终确定内容日历前，务必运行缺口报告
从不更新关键词研究	用户搜索行为会演变；2年前的查询意图或搜索量可能已变化	每年审核顶级内容；根据当前SERP数据更新关键词目标
忽略摘要和AI Overview展示情况	将所有关键词视为仅自然搜索渠道，会错过比自然搜索更容易取胜的AEO和GEO机会	研究期间记录每个关键词的SERP功能和AI Overview状态
将所有关键词视为仅自然搜索渠道	默认采用传统SEO，而不检查关键词是否更适合通过摘要或AI引用取胜	对所有优先级关键词进行三渠道评分；分配优先级渠道
对未触发AI Overview的关键词进行GEO评分	为AI引擎不生成答案的关键词分配GEO机会，会浪费精力	务必手动验证AI Overview展示情况；如果未触发AI回答，GEO评分=0
混淆研究与优化阶段	未完成关键词研究就尝试优化内容格式以获取摘要或AI引用，会导致过早决策	完成完整的研究流程（种子关键词>拓展>分类>评分>聚类）后，再进行任何优化工作

Gotchas

注意事项

Traffic potential and search volume diverge most on commercial keywords - A keyword with 300 monthly searches may be the primary term for a page that also ranks for 40 related variants, giving it 8,000 monthly visits. Conversely, a 10,000-volume head term may have traffic potential of only 6,000 because ranking page 1 earns only a small CTR share. Always pull traffic potential from your tool (Ahrefs "TP", Semrush "Traffic"), not raw volume.
SERP feature presence changes by location, device, and login state - An AI Overview you see in a logged-in US Chrome session may not fire in an incognito session or in another country. Always verify SERP features in incognito mode and, where relevant, from the target country using a VPN or tool like SERP API, before assigning AEO or GEO scores.
Keyword cannibalization diagnosis is wrong if you use position average - Google Search Console averages positions across all queries and dates. Two pages fighting for the same query may each show position 8 in the average, but the reality is one shows at 3 and the other at 15, alternating week by week. Filter GSC by specific queries and look for multiple pages appearing or for high impression/low click patterns on the same query across different pages.
Tri-surface scoring becomes meaningless if you score GEO for navigational queries - Navigational queries ("brand login", "product dashboard") almost never trigger AI Overviews. Assigning any GEO score above zero to navigational keywords inflates composite scores and misdirects content effort. GEO score must be zero unless you have verified an AI Overview fires for that query.
Keyword clusters built from tool "related keywords" lists ignore SERP overlap - Tool-suggested related keywords group by semantic similarity, not by whether Google actually returns the same URLs for both queries. Two semantically similar keywords may trigger completely different SERPs (different content types, different competition). Validate cluster membership by checking that 5+ of the top 10 results overlap between the keywords.

商业关键词的流量潜力与搜索量差异最大 - 一个月搜索量300次的关键词，对应的主页面可能覆盖40个相关变体，带来8000次月访问量。反之，一个月搜索量10000次的核心关键词，流量潜力可能仅为6000次，因为首页排名的点击率份额较低。始终从工具中获取流量潜力数据（Ahrefs的"TP"、Semrush的"Traffic"），而非原始搜索量。
SERP功能的存在情况因位置、设备和登录状态而异 - 你在登录状态下的美国Chrome会话中看到的AI Overview，在无痕会话或其他国家可能不会触发。分配AEO或GEO评分前，务必在无痕模式下验证SERP功能，必要时使用VPN或SERP API等工具从目标国家验证。
使用平均排名诊断关键词 cannibalization会出错 - Google Search Console会对所有查询和日期的排名取平均值。两个页面竞争同一查询时，可能各自显示平均排名8，但实际情况是一个页面排名3，另一个排名15，每周交替。按特定查询筛选GSC数据，查看是否有多个页面出现，或同一查询在不同页面上的高展示量、低点击率模式。
对导航型查询进行GEO评分毫无意义 - 导航型查询（如“品牌登录”、“产品仪表盘”）几乎不会触发AI Overview。为导航型关键词分配高于0的GEO评分会夸大综合得分，误导内容投入。除非已验证该查询会触发AI Overview，否则GEO评分必须为0。
工具“相关关键词”列表构建的关键词集群忽略SERP重叠 - 工具推荐的相关关键词按语义相似度分组，而非Google是否为两个查询返回相同的URL。两个语义相似的关键词可能触发完全不同的SERP（不同内容类型、不同竞争）。验证集群成员资格的方法是检查关键词之间的前10个结果是否有5个以上重叠。

References

参考资料

For detailed content on specific topics, read the relevant file from

references/

```
references/search-intent-mapping.md
```
- Deep dive into the four intent types, classification signals, intent-to-content-type matrix, intent-to-surface mapping for AEO and GEO, AI Overview intent patterns, and how to validate intent assumptions from SERP data. Load when classifying a keyword list or writing a content brief.
```
references/keyword-clustering.md
```
- Methods for clustering keywords (semantic, SERP-based, modifier-based, surface-aware), building pillar-and-spoke topic clusters, surface-aware cluster annotation, and tooling options. Load when building a cluster or planning a content architecture.
```
references/tri-surface-scoring.md
```
- Complete scoring rubrics for organic (0-10), AEO (0-10), and GEO (0-10) opportunity scores, composite scoring, business-goal weighting, three worked examples, and limitations. Load when scoring keywords or producing a tri-surface keyword research report.
```
references/tool-specific-workflows.md
```
- Step-by-step workflows for Ahrefs, Semrush, Google Search Console + free tools, and ChatGPT/Perplexity manual audits. Includes spreadsheet template with columns, formulas, and conditional formatting. Load when performing hands-on keyword research with specific tools.

Only load a references file if the current task requires deep detail on that topic.

如需特定主题的详细内容，请查看

references/

中的相关文件：

```
references/search-intent-mapping.md
```
- 深入介绍四类意图类型、分类信号、意图与内容类型矩阵、AEO和GEO的意图与渠道对应关系、AI Overview意图模式，以及如何通过SERP数据验证意图假设。分类关键词列表或撰写内容brief时加载此文件。
```
references/keyword-clustering.md
```
- 关键词聚类方法（语义聚类、基于SERP的聚类、基于修饰词的聚类、渠道感知聚类）、构建支柱与分支主题集群、渠道感知集群标注，以及工具选项。构建集群或规划内容架构时加载此文件。
```
references/tri-surface-scoring.md
```
- 自然搜索（0-10）、AEO（0-10）和GEO（0-10）机会评分的完整标准、综合评分、业务目标加权、三个示例和局限性。为关键词评分或制作三渠道关键词研究报告时加载此文件。
```
references/tool-specific-workflows.md
```
- Ahrefs、Semrush、Google Search Console+免费工具、ChatGPT/Perplexity手动审核的分步流程。包含带列、公式和条件格式的电子表格模板。使用特定工具进行实际关键词研究时加载此文件。

仅当当前任务需要该主题的详细内容时，才加载参考文件。

Companion check

配套Skill检查

On first activation of this skill in a conversation: check which companion skills are installed by running
ls ~/.claude/skills/ ~/.agent/skills/ ~/.agents/skills/ .claude/skills/ .agent/skills/ .agents/skills/ 2>/dev/null
. Compare the results against the
recommended_skills
field in this file's frontmatter. For any that are missing, mention them once and offer to install:
npx skills add AbsolutelySkilled/AbsolutelySkilled --skill <name>
Skip entirely if
recommended_skills
is empty or all companions are already installed.

在对话中首次激活本Skill时：通过运行
ls ~/.claude/skills/ ~/.agent/skills/ ~/.agents/skills/ .claude/skills/ .agent/skills/ .agents/skills/ 2>/dev/null
检查已安装的配套Skill。将结果与本文件前言中的
recommended_skills
字段对比。对于缺失的Skill，提及一次并提供安装命令：
npx skills add AbsolutelySkilled/AbsolutelySkilled --skill <name>
如果
recommended_skills
为空或所有配套Skill已安装，则跳过此步骤。