Welcome to AIKode
Making websites efficient for AI agents — and the everyone who uses them.
When users ask ChatGPT or Claude about your website, or to find current information do they give accurate answers?
Or outdated information from pages that changed months ago?
AIKode solves this.
We provide tools and protocols to make your web-site AI-friendly, so AI assistants find and give users the right answers, faster.
The Problem
AI assistants are becoming the primary way people find information. But current websites aren't optimized for AI consumption:
- AI agents load entire pages (thousands of tokens) to find one answer
- No way to know if content is current or stale
- AI often gives outdated or wrong information
- Slow, inefficient, expensive for everyone involved
Result: Your users get frustrated. Your great documentation goes unused.
The Solution
AIKode introduces a simple system: an /ai/ directory on your site that provides:
- Progressive disclosure — AI agents load only the detail level they need
- Verification timestamps — Know when content was last validated
- Efficient indexing — 80-90% token reduction vs. full page crawls
- AI-specific guidance — Teach AI agents how to use your site optimally
Result: Faster, more accurate answers. Better user experience. Lower AI costs.
How It Works
Your site gets an /ai/ directory with:
index.ai — Site catalog with progressive detail levels
skill.claude.ai — AI-specific optimization patterns
howto-ai.html — Instructions for AI agents
qset-spec.txt — Quick format reference
When an AI agent visits your site, it discovers the /ai/ directory and uses it to navigate efficiently.
Instead of loading full pages, it can now read only the summary information to find where it needs to look for more detail.
This is made possible with a progressive-depth date-reading method called Fathom:
The AI - Using Fathom:
- Scans FA (Fathom-AI) levels — ultra-compact keywords (~10 tokens each)
- Checks F1 (Title) for relevance (~20-50 tokens) or
- Checks F2 (Quick Summary) for relevance (~50-100 tokens)
- Loads F4 (Detail) only when needed (~500-1500 tokens)
- Loads F6 if extreme detail with references is needed.
Traditional approach: Load entire page (5,000 tokens)
AIKode approach: Scan + targeted load (500-1,000 tokens)
Savings: 80-90% reduction
See It In Action
Live example: setfiles.org/ai/(approve sites)
Visit the /ai/ directory to see the system in action. Then try asking Claude or ChatGPT about SET files — watch how it uses the optimized index for faster, more accurate answers.
Technology
AIKode is built on open protocols:
- Fathom Protocol(approve sites) — Progressive disclosure with FA + F1-F6 levels
- SET File Format(approve sites) — Human-readable, machine-parseable data structure
- AbSynth Protocol(approve sites) — Code metadata system (for developer documentation)
All protocols are open and free to use.
Feel free to support our work or
Use any of commercial software.
Get Started
There are three ways to use AIKode:
Do It Yourself (Free)
- Read the protocol documentation?
- Create your
/ai/ directory manually
- Deploy to your site
- Test with AI assistants
Perfect for: Technical users, small sites, learning the system
WordPress Plugin (Freemium)
- Free tier: Basic index generation, manual updates
- Pro tier ($99/year): Full automation, AI-assisted generation, analytics
- Agency tier ($299/year): Unlimited sites, white-label, API access
Perfect for: WordPress sites, agencies, ongoing maintenance
Done-For-You Service
We create and deploy your /ai/ directory for you.
- One-time setup: $2,500-5,000
- Includes custom optimization and training
- Perfect for: Large documentation sites, enterprise deployments
Contact us? for a quote.
Who Is This For?
Documentation Sites
- Technical documentation
- API references
- Developer guides
- Knowledge bases
SaaS Companies
- Product documentation
- Help centers
- Getting started guides
Content Platforms
- Educational sites
- Tutorial platforms
- How-to guides
Any site where AI assistants answer user questions
Benefits
For Users
- Faster, more accurate answers from AI assistants
- Current information (no outdated responses)
- Better overall experience
For Site Owners
- Better discoverability through AI agents
- Reduced server load (fewer full-page fetches)
- Analytics on what AI agents access
- Future-proof as AI becomes primary interface
For AI Agents
- 80-90% token reduction
- Clear freshness indicators
- Site-specific optimization patterns
- Efficient, accurate responses
Next Steps
Learn More
Get Started
Stay Updated
About
AIKode is part of a broader initiative to make the web more efficient for AI agents while maintaining human readability. Created by Kirk Siqveland, developer of the SET file format, Fathom Protocol, and EMU ecosystem.
Related Projects:
- SetFiles.org(approve sites) — Human-readable configuration format
- Fathom Protocol — Progressive disclosure system
- AbSynth Protocol — Code metadata system
- EMU Ecosystem — Embedded markup tools
Questions? Get in touch? — we'd love to hear from you.
Ready to optimize your site? Start here? →
