TL;DR: I built an open-source template for managing SEO projects with Claude Code. It includes slash commands for session management, data analysis, and content review. First results: 10+ comprehensive long-form articles—better quality than pre-agentic AI writers, cheaper than human copywriters. Free tools for position tracking. Get it on GitHub.
SEO success doesn't come from a single burst of effort. It's months of consistent work: researching keywords, writing articles, analyzing performance, adjusting strategy, and repeating. It's tedious and time-consuming work. For running businesses, it's usually outsourced to SEO professionals (and it isn't cheap). Most startups give up not because SEO doesn't work, but because the grind wears them down before results show.
What if AI could handle the repetitive parts while you focus on strategy and creativity?
For the past year, I've used Claude Code as an integrated part of my e-commerce SEO workflow - not for one-off tasks, but as a consistent assistant across months of work. First results: it's nearly free beyond the Claude Code subscription, 10+ articles published, monthly reporting reduced from 2 hours to 15 minutes, and a content audit process that now runs in seconds instead of days.
Here's what I learned, plus the open-source template that makes this approach reproducible.
The Reality of SEO Work
Let's be honest about what SEO actually involves:
Phase 1-3: Research (Weeks 1-4)
- Audit your current site
- Research 100-200 keywords
- Analyze 3-5 competitors
- Build a content strategy
Phase 4-6: Execution (Months 2-6)
- Write 5-10 articles per month
- Optimize existing pages
- Build internal links
- Submit to search engines
Phase 7-8: Maintenance (Ongoing)
- Track rankings weekly
- Analyze performance monthly
- Update underperforming content
- Build external links
Each phase has repetitive tasks. Keyword research means running the same queries across multiple tools. Writing follows templates. Performance analysis requires pulling data from the same sources and comparing against the same benchmarks.
This is where AI agents excel: consistent execution of well-defined tasks.
What "Agentic" Means Here
I'm not just talking about AI writing your articles (though it requires your input). I'm talking about AI handling the full workflow around content:
- Session management: Loading context at start, saving progress at end
- Data analysis: Processing Yandex Webmaster and Google Search Console exports
- Quality control: Running multiple critics on drafts before publishing
- Report generation: Creating monthly performance reports from templates
- Content audit: Scanning your site for gaps and opportunities
These tasks add up. Analyzing a month's search console data might take 30 minutes manually. With a well-crafted prompt, it takes seconds.
The Template Structure
Here's how I organize an AI-assisted SEO project:
.claude/
├── context/ # Current state
│ ├── active-context.md # This sprint's priorities
│ └── progress.md # What's done, what's next
├── memory/ # Long-term knowledge
│ └── memory-bank.md # Strategic insights
├── skills/ # Slash commands
│ ├── start-session/
│ ├── end-session/
│ ├── monthly-report/
│ └── analyze-webmaster/
└── workflows/ # Process docs
├── SEO_WORKFLOW.md
└── CONTENT_WRITING_GUIDE.md
scripts/
└── content_audit/ # Python utility
└── main.py
content/
├── drafts/
├── published/
└── calendars/
research/
├── keywords/
├── analytics/
└── content-audit/
The .claude/ directory is the key. It contains everything the AI needs to understand your project and execute tasks consistently.
Slash Commands (Skills)
Claude Code supports "skills" - custom prompts that can be invoked with /command. Here are the ones I use daily:
/start-session
Loads project context at the beginning of work:
- Reads active priorities
- Checks progress status
- Loads memory bank
- Provides status summary
/end-session
Saves progress and commits changes:
- Updates progress tracking
- Commits to git with detailed message
- Provides session summary
/analyze-webmaster
Processes Yandex Webmaster exports:
- Analyzes queries, pages, devices, geography
- Identifies growth opportunities
- Generates detailed report
/monthly-report
Generates monthly performance report:
- Creates template from structure
- Sets up data snapshot directory
- Provides checklist for data collection
/review-article
Runs multiple critics on content:
- SEO compliance check
- Language quality (native-level review)
- E-E-A-T signals
- User intent satisfaction
- Readability analysis
- Commercial integration
This last one is particularly useful. Instead of hoping your article is good, you get structured feedback from multiple perspectives before publishing.
Want to try it? Clone the template and run /start-session to see it in action.
The Content Audit Utility
Beyond Claude Code skills, the template includes a Python utility for site-wide content analysis:
cd scripts/content_audit
python main.py --full
This scans your sitemap, extracts content, and generates:
- Site-wide content inventory (CSV)
- Gap analysis (keywords you're missing)
- CTR optimization opportunities
- Cannibalization detection (pages competing for the same keywords)
The output integrates with Google Search Console and Yandex Webmaster data for a complete picture.
Required Input and How to Get It
This data could be obtained via API in future releases, but for now the workflow requires data from several sources:
| Data | Source | How to Get |
|---|---|---|
| Keywords | Yandex Wordstat, Ahrefs, SEMrush | Export to CSV |
| Search performance | Google Search Console, Yandex Webmaster | Download reports |
| Site content | Your sitemap | Automatic via Python script |
| Competitor data | Manual research | Document in research/ |
The template includes clear locations for each data type and instructions for processing.
What This Doesn't Do
Let's be honest about limitations:
- Won't write final articles without your input: AI drafts still need human input and polishing
- Won't guarantee rankings: SEO fundamentals still apply
- Won't replace expertise: You need to understand SEO to direct the AI
- Won't generate images for you: It will provide prompts for AI generation, but Claude lacks image generation
This is a productivity multiplier, not a magic solution.
Getting Started
1. Clone the Template
git clone https://github.com/stikhonchuk/ai-seo-workflow.git my-seo-project
cd my-seo-project
rm -rf .git && git init
2. Configure for Your Project
cp scripts/content_audit/config.example.py scripts/content_audit/config.py
# Edit config.py with your domain
3. Update Context Files
Replace placeholders in:
.claude/context/active-context.md.claude/context/progress.md.claude/memory/memory-bank.md
4. Start Using
claude # Start Claude Code
/start-session # Load context
That's it. The template handles the rest.
Multi-Project Architecture
If you want to contribute to the project, the template supports a hub-and-spoke model:
┌─────────────────────────────────────┐
│ PUBLIC TEMPLATE (ai-seo-workflow) │
│ - Generic workflows │
│ - Shared scripts │
│ - Community improvements │
└─────────────────────────────────────┘
▲ ▲
│ cherry-pick │ cherry-pick
│ (anonymized) │ (anonymized)
│ │
┌─────────────────┐ ┌─────────────────┐
│ project-a │ │ project-b │
│ (private) │ │ (private) │
└─────────────────┘ └─────────────────┘
Improvements flow both ways:
- Template updates → private projects via
git pull upstream - Private improvements → template via anonymize + cherry-pick
An anonymization script (scripts/anonymize.py) helps prepare files for public contribution by replacing project-specific data with placeholders.
Work in Progress
This template is actively evolving. Current areas of development:
- [ ] Better integration with Google Analytics 4
- [ ] Image generation prompt templates
- [ ] Link building workflow automation
- [ ] Multi-language content support
Contributing
If you try this workflow and improve it, contributions are welcome:
- Fork the repository
- Make your changes
- Run
python scripts/anonymize.py --check .to ensure no private data - Submit a pull request
Even bug reports and documentation improvements help.
Why Open Source This?
Three reasons:
- Forced better documentation: Knowing others will use it makes you document properly
- Community improvements: Others might solve problems I haven't encountered
- Reference implementation: Shows one way to structure AI-assisted workflows
The unique value isn't the SEO advice (plenty of that exists). It's the integration pattern: how to make AI a consistent part of a long-running project, not just a one-off tool.
FAQ
Do I need Claude Pro/Max subscription?
Claude Code requires subscription or API access, so you'll need an Anthropic API key with credits. The template itself is free.
Can I use this with ChatGPT or other AI tools?
The skills are written for Claude Code specifically, but the workflow structure (context files, memory bank, progress tracking) could be adapted to other tools.
What if I don't use Yandex Webmaster?
The template includes Google Search Console analysis too. Yandex-specific features are optional.
Repository: github.com/stikhonchuk/ai-seo-workflow
Built with: Claude Code by Anthropic
This is v1.0 of the template. I've been using variations of this workflow for a year. The public version is new and likely has rough edges. Feedback appreciated.
Top comments (0)