Skip to content

Experimental Projects

The experimental projects published by the Open LLMO Research Initiative. All ship in Draft / Experimental state. Formal specification status is deferred to Phase 3.

ProjectRoleAnalogStatus
1. LLMOFramework ScoreMeasure AI discoverability of a siteLighthouse ScoreIndicators being drafted (Draft v0.1 in Phase 1)
2. LLMOFramework BenchmarkExperimentally compare site structuresIndustry-standard benchmarkPlanning (Phase 1-2)
3. LLMOFramework CompatibleCertification badge for compliant sites”Certified” markRoadmap only (Phase 3)

Per-site score of how recognizable, citable, and parseable content is for AI. The AI-era counterpart to SEO’s Domain Authority or Lighthouse Score.

IndicatorDescription
Citation VisibilityWhether the content gets cited by AI
Chunk ReadabilityHow well the content chunks
Semantic StructureHow explicit the semantic structure is
AI CrawlabilityAI crawler compatibility
llms.txtllms.txt compliance
Markdown QualityStructural quality
Entity ClarityEase of entity recognition
Retrieval StabilityRetrieval consistency

Every indicator ships with a calculation formula and OSS checker code. Lighthouse earned trust because it was measurable and reproducible, and this project follows the same principle.

llmo-checker is planned for Phase 1.

npx llmo-checker https://example.com
LLMOFramework Score: 74
Citation Visibility: 81
Semantic Chunkability: 68
AI Readability: 77
Grounding Stability: 70

Indicator definitions are being drafted. Draft v0.1 publication is targeted for Phase 1 (timing TBD).


Experimental comparison of which site structures perform best for AI. No standard benchmark exists for AI retrieval and citation yet, so this project proposes a measurement methodology first.

  • Markdown vs HTML
  • FAQ schema presence
  • Table structure
  • Chunk size
  • Citation format
  • Internal linking
  • GitHub integration
  • llms.txt compliance
  • MCP exposure

Each experiment ships as a Reproducible Benchmark Report on GitHub and on this site, including the dataset, measurement scripts, raw results, and evaluation prompts.

Planning stage. The first comparison experiment (Markdown vs HTML retrieval efficiency) is planned for Phase 1.


Certification mark for sites that comply with AI-optimized structure. Intended for SaaS, documentation sites, OSS projects, and AI products to display.

[ LLMOFramework Compatible ]
[ AI Retrieval Ready ]
[ Grounding Optimized ]
RequirementContent
llms.txt placementA valid llms.txt exists at the site root
Semantic StructureMajor pages satisfy heading hierarchy and semantic HTML
Chunk OptimizationMajor sections fit within the recommended chunk size range
Grounding-friendly DocsCitations, data sources, and update dates are explicit

Roadmap only. Positioned at Phase 3 (last). The reasons:

  • Certification depends on ecosystem adoption, so Score and Benchmark must mature first
  • Issuing certification while solo-operated reads as authority cosplay and erodes trust
  • The Compatible badge will only be designed after the Open Source community has produced third-party adoption

PhaseProject progress
Phase 0 (current)Indicator drafting, project concept publication
Phase 1Score Draft v0.1, llmo-checker OSS, first Benchmark Report
Phase 2Score revision, continuous Benchmark updates, community feedback integration
Phase 3Compatible certification design, formal specifications, Working Group formation

Source code and discussion for each project are public at the GitHub repository and Issues.