SEO Article Generator

The most powerful open-source AI-powered SEO article generation pipeline.

Crawl thousands of articles from 5 search engines, extract clean content, and rewrite them into unique SEO-optimized articles using local AI — all from a single command.

Quick Start • Features • How It Works • Usage • Configuration

---

Why SEO Article Generator?

Most SEO content tools are expensive SaaS products with monthly fees, word limits, and vendor lock-in. SEO Article Generator is a free, open-source alternative that runs entirely on your machine using local AI models via Ollama. No API keys. No usage limits. No recurring costs.

• Scale: Generate hundreds or thousands of unique articles in a single run
• Quality: AI-rewritten content with natural keyword placement, proper headings, and SEO metadata
• Privacy: Everything runs locally — your content never leaves your machine
• Resilient: Built-in checkpoint system lets you pause and resume at any time
• Flexible: Works with any niche or topic, any Ollama-compatible model

---

Features

Multi-Engine Search Aggregation

Searches 5 engines simultaneously — DuckDuckGo, Bing, Google, Reddit, and Quora — with 25 query variations per keyword for maximum coverage. Automatic deduplication ensures no duplicate sources.

Intelligent Content Extraction

Uses trafilatura for state-of-the-art content extraction. Automatically strips navigation, ads, and boilerplate. Filters out junk content (Reddit drama, app store listings, etc.) with regex-based detection.

AI-Powered SEO Rewriting

Rewrites every article using a local LLM through Ollama. Each article gets:

• Completely rephrased content (unique enough to pass plagiarism checks)
• Natural keyword placement (3-5 mentions per article)
• SEO-optimized title and meta description
• Proper heading structure (## headings)
• URL-friendly slug
• Auto-categorization

Relevance Scoring & Filtering

Every extracted article is scored for relevance against your target keywords. Configurable thresholds let you control quality vs. quantity. Content is also filtered by minimum word count and deduplicated by title and content fingerprinting.

Checkpoint & Resume System

Long crawls are protected by automatic checkpoints saved every N articles. If your machine crashes, your internet drops, or you just need to stop — run with --resume to pick up exactly where you left off. No work is ever lost.

Batch Processing

Run 10+ keyword campaigns simultaneously with the batch runner. Processes jobs in parallel batches with per-job logging and comprehensive status reporting.

Excel Export with Formatting

Exports to professionally formatted .xlsx files with:

• Frozen header rows and auto-filters
• Color-coded rewrite status (green/yellow/red)
• Summary sheet with statistics, top sources, and category breakdown
• 15 data columns including original and rewritten content

Blog Integration

Built-in exporter converts articles to JavaScript blog post objects with HTML content, table of contents, reading time, categories, and tags — ready to drop into any static site or CMS.

---

How It Works

┌─────────────────────────────────────────────────────────────┐
│                    SEO Article Generator                     │
├──────────┬──────────┬──────────────┬──────────┬─────────────┤
│  SEARCH  │ EXTRACT  │   FILTER     │ REWRITE  │   EXPORT    │
│          │          │              │          │             │
│ DuckDuck │ Fetch    │ Word count   │ Ollama   │ Excel       │
│ Bing     │ HTML     │ Relevance    │ LLM      │ Blog JS     │
│ Google   │ Clean    │ Dedup        │ SEO      │ Stats       │
│ Reddit   │ Content  │ Junk filter  │ Meta     │ Summary     │
│ Quora    │ Metadata │ Domain block │ Slug     │             │
└──────────┴──────────┴──────────────┴──────────┴─────────────┘
     │           │            │            │           │
     ▼           ▼            ▼            ▼           ▼
  5 engines   trafilatura  Smart       Local AI    Formatted
  25 queries  extraction   filtering   rewriting   .xlsx output
  per keyword              pipeline    via Ollama

---

Quick Start

Prerequisites

• Python 3.8+
• Ollama installed and running (for AI rewriting)

Installation

# Clone the repository
git clone https://github.com/aymenhmaidiwastaken/seo-article-generator.git
cd seo-article-generator

# Install dependencies
pip install -r requirements.txt

# Pull an AI model (llama3 recommended)
ollama pull llama3

Generate Your First Articles

# Generate articles about any topic
python -m article_crawler "keto diet" "keto recipes" "ketogenic lifestyle"

That's it. The tool will search the web, extract articles, rewrite them with AI, and export everything to articles_output.xlsx.

---

Usage

Basic Usage

# Single keyword
python -m article_crawler "mobile app development"

# Multiple keywords for broader coverage
python -m article_crawler "react native" "flutter" "cross platform apps"

# Custom output file and target count
python -m article_crawler "machine learning" -o ml_articles.xlsx -t 500

Advanced Usage

# Crawl without AI rewriting (just collect raw articles)
python -m article_crawler "python programming" --no-rewrite

# Resume an interrupted crawl
python -m article_crawler "python programming" --resume

# Rewrite articles from an existing Excel file
python -m article_crawler --rewrite-only raw_articles.xlsx -o rewritten.xlsx

# Use specific search engines only
python -m article_crawler "web3" --engines duckduckgo bing

# Use a different AI model
python -m article_crawler "fitness tips" --model mistral

# Fine-tune extraction filters
python -m article_crawler "seo tips" --min-words 200 --min-relevance 0.05

# Verbose logging for debugging
python -m article_crawler "blockchain" -v

Batch Processing

For large-scale content generation across multiple topics:

# Edit resume_all.py with your keyword campaigns, then:
python resume_all.py

This processes multiple keyword sets in parallel batches with automatic logging.

Blog Export

Convert generated articles to blog-ready JavaScript:

python export_to_blog.py --input articles_output.xlsx --limit 100

---

Configuration

CLI Options

Option	Default	Description
keywords	(required)	Search keywords/phrases
-o, --output	articles_output.xlsx	Output Excel file path
-t, --target	5000	Target number of articles
--min-words	300	Minimum article word count
--min-relevance	0.1	Relevance score threshold (0-1)
--engines	duckduckgo bing reddit quora	Search engines to use
--model	llama3	Ollama model for rewriting
--ollama-url	http://localhost:11434	Ollama server URL
--resume	false	Resume from checkpoint
--no-rewrite	false	Skip AI rewriting phase
--rewrite-only	—	Rewrite existing Excel file
--skip-dedup	false	Allow duplicate articles
--checkpoint-every	25	Save progress every N articles
-v, --verbose	false	Enable debug logging

Supported Search Engines

Engine	Method	Reliability	Notes
DuckDuckGo	API (ddgs)	High	Most reliable, recommended
Bing	Web scraping	High	Good complement to DDG
Google	Web scraping	Low	Frequently rate-limited
Reddit	JSON API	Medium	Finds articles shared on Reddit
Quora	DDG site search	Medium	Q&A content via DuckDuckGo

Supported AI Models

Any model available through Ollama works. Recommended:

Model	Speed	Quality	Best For
llama3	Fast	High	General-purpose rewriting
llama3:70b	Slow	Very High	Premium quality output
mistral	Fast	Good	Quick drafts
mixtral	Medium	High	Balanced speed/quality
gemma2	Fast	Good	Lightweight rewriting

---

Project Structure

seo-article-generator/
├── article_crawler/          # Core package
│   ├── __init__.py           # Package metadata
│   ├── __main__.py           # Module entry point
│   ├── main.py               # CLI & orchestrator
│   ├── searcher.py           # Multi-engine search (5 engines)
│   ├── extractor.py          # Content extraction & filtering
│   ├── rewriter.py           # AI rewriting via Ollama
│   ├── exporter.py           # Excel export with formatting
│   ├── checkpoint.py         # Save/resume system
│   └── utils.py              # Helpers (delays, slugs, URL filtering)
├── run.py                    # Convenience wrapper
├── resume_all.py             # Batch job runner
├── export_to_blog.py         # Blog CMS exporter
├── requirements.txt          # Python dependencies
└── README.md

---

Output Format

The generated Excel file contains 15 columns:

Column	Description
Title	Original article title (cleaned)
Article Content	Original extracted content
Source URL	Where the article was found
Date Published	Publication date
Author	Article author
Word Count	Number of words
Meta Description	SEO meta description
Keywords Found	Matching target keywords
Headings	H1-H3 heading structure
Source Domain	Website domain
Rewritten Title	AI-generated SEO title
Rewritten Content	Fully rewritten article
Rewrite Status	completed / pending / failed
Slug	URL-friendly slug
Category	Auto-assigned category

A Summary sheet is also generated with statistics, top source domains, and category distribution.

---

Tips for Best Results

1. Use 3-5 specific keywords per run for focused, high-relevance articles
2. Start with --no-rewrite to preview what gets collected before spending time on AI rewriting
3. Use --resume liberally — the checkpoint system makes it safe to stop and restart anytime
4. Lower --min-relevance to 0.05 if you want more articles (at the cost of some off-topic results)
5. Use llama3:70b for the highest quality rewrites if you have the hardware
6. Run batch jobs overnight using resume_all.py for large-scale content generation

---

Contributing

Contributions are welcome! Some ideas for improvement:

• Add more search engines (Yandex, Brave Search, etc.)
• Support cloud LLM APIs (OpenAI, Anthropic, etc.) as rewriting backends
• Add Markdown/HTML export formats
• Implement content spinning with synonym replacement
• Add image extraction and AI image generation
• Build a web UI dashboard for monitoring crawl progress
• WordPress/Ghost/Hugo direct publishing integration

---

License

This project is licensed under the MIT License — see the LICENSE file for details.

---

_{Built for content creators, SEO professionals, and developers who believe great tools should be free and open.}