b2b-seo-content-writer
End-to-end B2B SEO article production following a 12-step SOP. Covers 8 article cluster types (educational, buying guide, comparison, specifications, application scenarios, process/quality, certificat
SKILL.mdskills/b2b-seo-content-writer/SKILL.md
GitHubB2B SEO Content Writer
12-step SOP for producing high-conversion B2B SEO articles with EEAT, GEO, and Schema compliance.
Core Writing Principles
1. SEO-Optimized, User-First Content
Every article must serve two masters equally: search engines and human readers.
- Search Intent First: Before writing, identify what the searcher actually wants when they type the keyword. A "how to choose" query wants decision criteria, not a product pitch. A "what is" query wants definitions and basics, not a buying guide.
- Topic Coverage: Address the questions users naturally have after searching this keyword. Cover related subtopics that Google's "People Also Ask" and related searches surface.
- Information Density: Every paragraph delivers actionable information. If a paragraph can be deleted without losing value, delete it.
- Rich Yet Readable: Use bullet lists, comparison tables, and structured headings to make complex B2B information scannable. Short paragraphs. Clear hierarchy.
- Google Best Practices: Primary keyword in title, H1, first 100 words, one H2, meta description, and URL slug. Secondary keywords distributed naturally. No stuffing.
2. Bilingual Output Rule
- All interactions with the user: Use Chinese for questions, explanations, confirmations, and status updates.
- All generated content (deliverables): Produce English first, then immediately follow with the Chinese translation. One block at a time: English block → Chinese block → next English block → next Chinese block.
- Image file names and Alt text: English only.
- Schema markup: English only.
3. No AI-Sounding Content, No Exaggerated Claims
The article must read like it was written by a factory engineer — not a marketing AI.
AI Patterns to Eliminate:
| AI Pattern | Example | Why It's Bad |
|---|---|---|
| Overly polished transitions | "Furthermore, it is worth noting that..." | Reads like GPT, not a human |
| Vague superlatives | "industry-leading", "state-of-the-art", "world-class", "unparalleled" | Empty words that fit any product |
| Fake urgency / hype | "Don't miss out!", "Game-changing!", "Revolutionary!" | Damages credibility with B2B buyers |
| Generic closing paragraphs | "In conclusion, choosing the right X is essential for your business success." | Says nothing, wastes reader's time |
| Over-explaining the obvious | "A warehouse is a place where goods are stored..." | Insults the reader's intelligence |
| Symmetrical, balanced sentence structures throughout | Every paragraph has exactly 2-3 sentences of similar length | Classic GPT fingerprint |
| Repeated sentence starters | Consecutive sentences starting with "This", "It", "The" | Monotonous and detectable |
Exaggerated Claims to Avoid:
| Don't Say | Problem |
|---|---|
| "the best scissor lift on the market" | Cannot be verified, damages trust |
| "guaranteed to increase efficiency by 50%" | Fake precision, no source |
| "trusted by thousands of customers worldwide" | Vague number, sounds made up |
| "never requires maintenance" | Physically impossible, dishonest |
| "solves all your material handling problems" | Over-promising, loses credibility |
The Rule: Write like an engineer explaining their work to a peer. Be specific. Be honest. If a statement can''t be backed by a measurement, a test report, or a customer reference, don''t write it. Under-promise, over-deliver.
4. International Website Standards
This is an international-facing website. All content must follow international conventions:
- ✅ English punctuation only: Use
.,:;?!-—'"— never,。:;""''even in Chinese translations. - ✅ Metric units with imperial in parentheses where helpful: "5000 kg (11,023 lbs)".
- ✅ ISO date format: YYYY-MM-DD.
- ✅ 24-hour time or "AM/PM" — never Chinese time notation.
- ✅ No region-specific references unless explicitly about that region (e.g., don't say "在我们国内" — say "in the Chinese market").
- ✅ Currency: Use USD/EUR or specify the currency explicitly. Don't assume RMB.
Prerequisites
Before starting, collect from the user:
- Company name, website domain, and industry/niche
- Product/service categories (for internal linking targets)
- Target audience (job titles, industries, use cases)
- Existing brand assets: factory details, certifications, case studies, project photos
- SEO tool access: SEMrush or equivalent (for keyword data)
- Competitor URLs (3-5 competing sites to analyze)
If any prerequisite is missing, ask the user before proceeding.
Article Cluster Types
Always begin by identifying which of the 8 cluster types the article belongs to:
| Cluster | Core Question | Typical Titles | Funnel | Value |
|---|---|---|---|---|
| Educational | What is X? | What is X, X explained, X types, X uses | TOFU | Traffic entry |
| Buying Guide | How do I choose? | How to choose, selection guide, best X for Y | MOFU | Very high |
| Comparison | X vs Y? | X vs Y, compare, pros and cons, alternative | MOFU | Very high |
| Specifications | What are the specs? | Specifications, dimensions, standards, testing | TOFU/MOFU | Strong |
| Application Scenarios | Who/what project is it for? | Best for hotels, for retailers, for hospitals | MOFU/BOFU | Strongest for B2B |
| Process/Quality | How? Why problems? | Manufacturing process, QC, troubleshooting, defects | TOFU/MOFU | Strong |
| Certification/Compliance | What certifications? | Certification explained, CE, REACH, ISO, how to pass | MOFU/BOFU | Medium-high |
| Procurement/Supplier | How to work with supplier? | OEM vs ODM, MOQ, lead time, factory audit checklist | MOFU/BOFU | Highest (closest to inquiry) |
The 12-Step Workflow
Execute steps in order. Do not skip steps.
Step 1: Cluster Title Research
- Plan 5-10 article titles for the target cluster type
- Use the cluster table above to guide title directions
- Output a bullet list of candidate titles with target keyword for each
Step 2: Competitor Outline → Own Outline
- Analyze top 3-5 ranking pages for the primary keyword
- Extract their H2/H3 structure
- Build a superior outline that:
- Covers what competitors cover
- Adds unique angles (factory perspective, engineer insights, real project data)
- Includes an FAQ section (4-6 questions)
- Output: complete article outline with H2/H3 headings
Step 3: SEMrush Keyword Mining
- Input primary keyword into SEMrush
- Extract:
- Primary keyword (search volume, difficulty)
- 5-10 secondary keywords (long-tail variations)
- 5-10 LSI/semantic keywords
- Log all keywords with search volume and intent type
Step 4: GPT-Generated EEAT Material
Generate EEAT-enhancing content from the company's unique perspective:
- Expertise: 2-3 paragraphs from an engineer or technical expert viewpoint (real process details, parameter explanations, material knowledge)
- Experience: 1-2 real project case summaries (client name/industry, problem, solution, result — keep anonymous if needed)
- Authoritativeness: Cite industry standards (ISO, ASTM, CE, REACH), reference authoritative sources
- Trustworthiness: Factory photos, QC process, certification list, warranty/after-sales info
Step 5: Internal Link + External Reference Plan
External Links (3-4 minimum):
- 1 industry authority/media source
- 1 peer/upstream-downstream site
- 1 Wikipedia reference
- 1 international well-known case/brand
Internal Links (3 minimum per article):
- 1 product page link (anchor: product name keyword)
- 1 case study page link (anchor: industry + scenario keyword)
- 1 contact page link (anchor: CTA keyword)
- Additional: related cluster articles (anchor: related topic keyword)
Output a link plan table with: anchor text, target URL, link type (internal/external), placement context.
Step 6: Submit Outline to AI
Provide the AI writer with:
- Complete article outline (from Step 2)
- Target audience description
- Brand/company context
- Tone of voice guidelines
Step 7: Keyword Placement + EEAT Markup Instructions
Provide the AI with precise placement rules:
- Primary keyword: in Title, H1, first paragraph, one H2, Meta Description
- Secondary keywords: distributed across H2/H3 headings
- LSI keywords: naturally woven into body text
- EEAT markers: insert at designated positions in the outline:
- [EXPERTISE] after technical explanations
- [EXPERIENCE] in application scenario sections
- [AUTHORITY] near standard/certification mentions
- [TRUST] in purchasing guide or supplier sections
Keyword density: primary keyword 0.5-1.5%, no stuffing.
Step 8: GEO Article Generation
Produce the final article with these requirements:
- Length by Funnel Stage:
Funnel Cluster Types Minimum Words TOFU (Top of Funnel) Educational, Specifications 1,500 MOFU (Middle of Funnel) Buying Guide, Comparison, Process/Quality, Certification 2,500 BOFU (Bottom of Funnel) Application Scenarios, Procurement/Supplier 3,000 - GEO summaries: Include ≥ 3 of the following:
- TL;DR (50-80 word executive summary at top)
- Key Takeaways (3-5 bullet points)
- Comparison Table (when applicable)
- Step-by-Step List (for how-to content)
- FAQ (4-6 Q&A at end)
- Structure rules:
- First paragraph ≤ 100 words, direct to the point, include answer summary
- Each H2 section ≥ 200 words
- No cliché openings ("In summary", "To conclude", etc.)
- Data and facts must cite sources
- No fabricated data: Never invent statistics, client names, project metrics, or certifications. Use [placeholder] markers when real data is unavailable from the user.
- EEAT markers inserted per Step 7 positions
- Internal/external links inserted per Step 5 plan Language Support:
- Generate the article in the language requested by the user (English default)
- If user requests translation, produce a full translated version
- Maintain all internal link targets, Alt text, and schema JSON in both languages
- SEO keywords should be researched separately for each target language
Step 9: WordPress Preview QA
Quality checklist before publishing:
- Word count ≥ 2500 (verify with word counter)
- EEAT content correctly inserted at designated positions
- ≥ 3 GEO summary types included
- Primary keyword in: Title, H1, first paragraph, Meta Title
- H1/H2/H3 hierarchy correct (no skipped levels)
- No keyword stuffing
- Mobile layout acceptable
- Image Alt text filled
- Meta Description ≤ 160 characters
Step 9.5: Content Quality Audit
Before finalizing the article, run a quality audit across 5 dimensions:
1. AI-Generated Content Detection
Scan for overused AI patterns and virtual/fabricated information:
- Cliché openings: Check for "In today's fast-paced world", "In the ever-evolving landscape of", "In conclusion", "To summarize"
- Fluff phrases: "It is important to note that", "It is worth mentioning that", "As previously stated"
- Overly generic claims: "Industry-leading", "State-of-the-art", "World-class" (unless backed by specific evidence)
- Fabricated data: Flag any statistic, percentage, or "studies show" claim that lacks a source citation
- Virtual/fake case studies: Verify that named clients, project locations, and specific numbers can be traced back to real data
Action: Replace flagged content. If a claim cannot be verified, either add a source or remove it.
2. Factual Verification
- Every data point has a source (linked or cited)
- Technical specifications match Gradin product documentation
- Certification names and numbers are accurate (ISO, ANSI, CE)
- Client case details are factual (dates, locations, metrics)
3. External Link Health Check
For each external link from Step 5:
- Run a GET request — confirm HTTP status is 2xx or 3xx
- Check the linked page is relevant to the anchor text
- Verify the link target still exists (Wikipedia pages, MHI resources, OSHA pages)
- Flag 404 / 403 / 5xx results for replacement
4. Internal Link Validation
For each internal link from Step 5:
- URL path matches the site's actual URL structure
- Slug format is correct (hyphen-case, no trailing slashes)
- Anchor text is descriptive (not "click here" or "read more")
5. Readability Assessment
- Average sentence ≤ 25 words (spot-check 5 random sentences)
- Paragraph length ≤ 4 sentences
- No single sentence exceeds 40 words
- Flesch Reading Ease score ≥ 50 (target for B2B audience)
Output: Pass/Fail report for each dimension with specific flagged items and suggested fixes.
Step 10: Images + External Links + Internal Links
Images:
Minimum 1 image per 500 words. For each image, output a placement table with exact paragraph insertion points.
Image Placement Table
For each image, specify:
| # | File Name | Type | Insert After (Paragraph) | Alt Text |
|---|---|---|---|---|
| 1 | keyword-description.jpg |
[product / process / chart / case / infographic] | [H2 section name, sub-paragraph] | [Descriptive Alt with keyword] |
Image Types by Cluster
| Cluster Type | Recommended Image Types |
|---|---|
| Educational / Specifications | Product photos, dimension diagrams, cutaway views |
| Buying Guide / Comparison | Comparison infographics, feature callout diagrams |
| Application Scenarios | Customer site photos, before/after installation |
| Process/Quality | Factory floor photos, QC testing equipment, weld inspection |
| Certification/Compliance | Certification badges, testing lab photos, standard logos |
| Procurement/Supplier | Factory exterior, team photos, packaging/loading dock |
Dual-Path Image Sourcing:
Path A — Local images exist: Use real Gradin photos directly. Follow file naming convention keyword-description.jpg. All images get descriptive Alt text (include keyword when natural).
Path B — Local images not available: Provide AI image generation prompts for each missing image. Prompt format:
[Style descriptor] image of [subject] in [context], [key visible elements], [lighting/mood], [technical specification] --ar [ratio]
External Links:
- Insert per Step 5 plan at corresponding text positions
- Attribute:
target="_blank" rel="noopener noreferrer" - Authority sources do not need nofollow
Internal Links:
- Insert per Step 5 plan
- Ensure minimum: 1 product page + 1 case page + 1 contact page
Step 11: Schema Markup
Add structured data. Required schemas:
Article Schema (all articles):
{
"@context": "https://schema.org",
"@type": "Article",
"headline": "[Title]",
"author": { "@type": "Organization", "name": "[Company]" },
"datePublished": "[Date]",
"dateModified": "[Date]",
"publisher": {
"@type": "Organization",
"name": "[Company]",
"logo": { "@type": "ImageObject", "url": "[Logo URL]" }
}
}
FAQ Schema (if article has FAQ section):
{
"@context": "https://schema.org",
"@type": "FAQPage",
"mainEntity": [
{ "@type": "Question", "name": "[Q]", "acceptedAnswer": { "@type": "Answer", "text": "[A]" } }
]
}
HowTo Schema (if article has step-by-step instructions):
{
"@context": "https://schema.org",
"@type": "HowTo",
"name": "[Title]",
"step": [{ "@type": "HowToStep", "name": "Step N", "text": "[Content]" }]
}
ImageObject Schema (for key images):
{
"@context": "https://schema.org",
"@type": "ImageObject",
"url": "[URL]",
"description": "[Description]",
"width": "[W]",
"height": "[H]"
}
Step 12: Keyword Record Archive
Log all article data into the tracking CSV (use template from assets/keyword-record-template.csv).
Fields to record: Article Title, Outline Summary, Core Keyword, Search Intent, Monthly Search Volume, Product Page Anchor Text & URL, Case Page Anchor Text & URL, Contact Page Anchor Text & URL, Cluster Article Anchor Text & URL, External Link 1-4 (Type/Source/URL), Article URL, Publish Date, Word Count, EEAT Markers (✓/✗), GEO Summary Types, Schema Markup Types, Index Status.
Workflow Summary
Step 1 Cluster Title Research
↓
Step 2 Competitor Outline → Own Outline
↓
Step 3 SEMrush Keyword Mining
↓
Step 4 GPT EEAT Material (factory/project perspective)
↓
Step 5 Internal Link Plan + External Reference Sources
↓
Step 6 Submit Outline to AI
↓
Step 7 Keyword Placement + EEAT Markup Instructions
↓
Step 8 GEO Article Generation (≥2500 words, ≥3 summary types)
↓
Step 9 WordPress Preview QA
↓
Step 9.5 Content Quality Audit (AI detection + link check)
↓
Step 10 Images + External Links + Internal Links
↓
Step 11 Schema Markup (Article/FAQ/HowTo/Image)
↓
Step 12 Keyword Record Archive
Execution Notes
- For each step, output the deliverable before moving to the next step. Do not batch multiple steps silently.
- When user confirms a step's output, proceed to the next.
- If any step requires data the user hasn't provided, pause and ask.
- Use the cluster type table to guide title research and outline structure.
- Reference
assets/article-cluster-research-template.mdfor Step 1 title research. - Reference
assets/keyword-record-template.csvfor Step 12 data logging.