Keyword Research: The Complete Process (From Requirements to Execution)

Chapter 2: Search Fundamentals · Question 20

Keyword research is not "coming up with a few words off the top of your head"—it is a systematic process that starts from business requirements, validates with data, and ultimately lands as a content plan. Doing keyword research well is equivalent to drawing a roadmap for your entire SEO strategy.

Why is keyword research the starting point of SEO?

• It determines what content you need to write
• It determines who your content can reach
• It determines how high your traffic ceiling is

Starting to write content without keyword research is like hitting the road without a map—you may go far, but in the wrong direction.

The Complete Process: 6 Steps

Step 1: Clarify business goals and user personas

Before opening any tool, answer these questions first:

1. What problem does my product/service solve?
2. Who are my target users? How do they describe their needs?
3. What will users search for before purchasing/deciding?
4. What do I want to achieve through search traffic? (Customer acquisition, brand awareness, market education?)

Step 2: Build a seed keyword library

Collect initial keywords from the following channels:

The goal of this step is "broad coverage and large volume"—do not filter yet; collect 50-200 seed keywords.

Step 3: Expand and quantify with tools

Input seed keywords into professional tools to obtain data such as search volume, competition level, and related long-tail keywords.

Recommended tool combinations:

Step 4: Classify and group

Organize the collected keywords by the following dimensions:

and group them bysearch intentClassification:

Take songbing.com.cn (a content site focused on Chinese organic marketing / AI searchSEO) as an example:

• Informational: users want to learn knowledge (e.g., "what is GEO", “what is organic traffic”) → corresponds to on-site what-is-geo/、what-is-organic-marketing/ Q&A pages like these
• Navigational: users are looking for a specific website/brand (e.g., "songbing 100 questions", “songbing.com.cn”) → corresponds to the homepage and directory pages
• Transactional: users are ready to act (e.g., "organic marketing training", “SEO managed services”) → if service pages are built in the future, use service pages to capture these
• Investigational: Users are in comparison-decision mode (e.g., "SEO vs GEO", “What's the difference between AEO and GEO”) → map to ai-summary-vs-traditional-ranking/、what-is-geo/ this type of comparison Q&A

Layer by business priority:

1. core keywords(directly tied to conversion) → prioritize homepage/product pages to capture
2. secondary keywords(indirectly tied) → use category pages and topic pages to capture
3. long-tail keyword(traffic supplement) → cover one by one with blog articles

Step 5: Assess difficulty and set priorities

Not every keyword is worth pursuing right away. Evaluate using three metrics:

Prioritize: keywords with moderate search volume + low competition + high business relevance. These are often long-tail keywords or niche-scenario keywords.

Step 6: Turn into a content plan

The end goal of keyword research is not an Excel sheet, but anexecutable content schedule：

1. Each target keyword maps to one piece/group of content
2. Specify the page type that captures each piece of content (homepage/category page/article page)
3. Set publish dates and priorities
4. PlanInternal linksrelationships (which articles link to each other, which point to core pages)

A real case: how songbing.com.cn does it

Take songbing.com.cn as an example. Its keyword research for the "Chinese organic marketing" topic follows this approach:

种子词：非付费营销、AI搜索优化、GEO
  → 扩展（用5118/百度指数挖长尾）：
     什么是GEO、GEO和SEO的区别、AI搜索怎么优化、
     怎么让AI推荐我的品牌、小红书搜索排名怎么做、
     新闻稿营销技巧、内容去重怎么处理...
  → 分类（按意图）：
     - 信息型（什么是GEO/什么是自然流量）→ 写成"XX是什么"问答页
     - 调研型（GEO和SEO区别/AEO是什么）  → 写成对比型问答页
     - 实操型（怎么让AI推荐我/怎么做小红书排名）→ 写成步骤型教程页
  → 落地：101 篇问答按 6 个章节组织，全部汇聚在 /100wen/ 目录页

This is the complete real-site walkthrough of the "6-step process" above: seed keywords expanded into 101 question-type keywords → each question maps to one Q&A → organized by chapter → aggregated on a directory page. You can also see that every article title on the site is itself a keyword (e.g., "What Is GEO?", "The Correct Use of Content Deduplication and Canonical Tags").

Advanced supplement: a practical 3-step keyword library build + 4-layer demand segmentation method

The methodology below comes from a hands-on breakdown on the author's WeChat Official Account. It is more "engineering-oriented" than the general process above, and is especially suited to content sites building a keyword library from scratch:

3 steps to build a keyword library:

1. Multi-channel keyword expansion (start with 500+): Use Baidu Index to see trends, 5118 to mine long-tail keywords, and site owner tools to check industry keywords; analyze the traffic keywords of the top 3 competitors; crawlZhihu/Tieba high-frequency questions to supplement "user's own words"-type demand
2. Data cleaning: Remove obscure terms with monthly search volume below 50, highly competitive terms whose first page is dominated by major sites, and terms unrelated to the business; retain 300-400 effective terms that are "easy to rank for and high-converting"
3. Classify and document: Use a spreadsheet to tag each term with "search intent + landing page type" to form the master keyword list

4-tier demand segmentation (corresponding to 4 page types, using songbing.com.cn as an example):

Key point:The decision stage has the highest conversion rate(Organic search conversion rate is about 5%-8%), prioritize optimizing this layer when the budget is limited; the repurchase layer has a higher conversion rate but requires long-term operation.

Source note: This section is consolidated from the author's WeChat Official Account article "The Full SEO Keyword Analysis Workflow | A Practical Guide from Keyword Expansion to Ranking" (2025-08-18); the original link is at the end of this article.

Practical recommendations

• Update the keyword library once per quarter: Search demand changes with the market; new terms keep emerging and old terms may become outdated. Set a fixed cadence (e.g., the start of each quarter) to rerun the data and capture new opportunities.
• Don't just look at search volume—look even more at "winnability": A term searched 100,000 times a month but whose top 10 are all big sites is worse than a term searched 500 times a month with zero competition (the "100,000 times/500 times" above are illustrative numbers, not real statistics). New sites especially should start with low-competition long-tail keywords.
• Keyword research must be coordinated with the content team: The research results must be turned into a content brief handed to writers, including target keywords, search intent, reference competitor content, and expected content structure; otherwise, no matter how good the research is, it won't be implemented.

Content sources and WeChat Official Account

Parts of this article are compiled from the author's WeChat Official Account articles:

• "The Full SEO Keyword Analysis Workflow | A Practical Guide from Keyword Expansion to Ranking"（2025-08-18）Original link

For more practical content on brand marketing, SEO, and AI search, follow the WeChat Official Account "AI Growth".