The SEO Engine That Runs Without Me (Or Any Tokens)
The SEO Engine That Runs Without Me (Or Any Tokens)
Most SEO tools charge $100+/month. Most AI SEO tools burn tokens on every run. This one costs zero dollars and zero tokens.
The SEO Content Engine is an 8-stage pipeline that goes from keyword to published draft without touching an LLM — unless you want it to. It uses DuckDuckGo suggest for keyword research, scrapes SERPs for competitor analysis, and templates for outlines. The LLM is a polish pass, not the engine.
The pipeline
Stage 1: Keyword discovery. Feed it a seed topic. It queries DuckDuckGo Suggest to find real search queries. No API key. No cost. Just HTTP requests.
Stage 2: Keyword filtering. It scores keywords by search volume signals, competition signals, and relevance to the seed topic. Low-quality keywords get filtered out before anything expensive happens.
Stage 3: SERP analysis. For the top keywords, it scrapes the first page of Google results. It extracts titles, meta descriptions, and content structure from ranking pages. This tells you what Google thinks the intent is.
Stage 4: Content outline. Based on SERP analysis, it generates a structured outline with H2s and H3s. The outline matches the content depth of top-ranking pages.
Stage 5: Draft generation. Template-driven first draft. For each section in the outline, it pulls in relevant data from SERP analysis. The draft is rough but structurally complete.
Stage 6: Internal linking. It scans existing content on your site and suggests internal links with anchor text.
Stage 7: SEO optimization. Title tag, meta description, URL slug, header hierarchy, keyword density check, and readability score.
Stage 8: Schema markup. Generates Article schema, FAQ schema if applicable, and breadcrumb schema.
Optional Stage 9: LLM polish pass (DeepSeek, ~$0.01 per article) for tone and fluency.
The template system
The engine works because the templates are good. I spent a week writing them.
Each content type — how-to, listicle, deep dive, comparison, news analysis — has a template with pre-written section structures, transition phrases, and CTA patterns. The template does 80% of the work. The SERP data fills in the specifics.
This is the insight that makes the system work: you don't need an LLM to write if your templates are good enough. SERP analysis tells you what to say. Templates tell you how to structure it. The LLM is just for the last 10% of polish.
What 10,000 pages would cost
Traditional approach:
- Ahrefs/SEMrush: $100-200/month
- Writer: $50-200/article
- Editor: $30-100/article
- Total for 10,000 pages: $800,000-$3,000,000
This engine:
- Keyword research: $0 (DuckDuckGo Suggest)
- SERP analysis: $0 (self-hosted scraping)
- Templates: $0 (built once)
- LLM polish (optional): ~$0.01/article
- Total for 10,000 pages: $0-$100
The bottleneck isn't cost. It's quality control. But when your cost per page is effectively zero, you can afford to be selective.
The architecture
- 8-stage pipeline: Keyword → SERP → Outline → Draft → Linking → SEO → Schema → Publish
- Template-driven: Content types, not LLM generations
- Free keyword data: DuckDuckGo Suggest, no API keys
- Self-hosted scraping: No third-party APIs for SERP data
- CMS adapters: WordPress, Ghost, Webflow, Contentful, Sanity
- LLM: Optional. DeepSeek when enabled. Silent fallback to template when key is missing.
The engine outputs markdown files. The CMS adapter pushes them to your platform. The only manual step is review. Everything else runs on cron.