geo-optimizer-skill: GEO Optimizer: The Open-Source Lighthouse for AI Citability

It audits robots.txt, llms.txt, schema, and brand signals, then turns research on generative engine optimization into a tool an AI agent can actually use.

8 min read • View on GitHub • More from auriti-labs

A technician studies a machine made of web standards instead of gears. A robots.txt scroll, an llms.txt ledger, and smaller modules for schema, content, and brand feed into a single lens marked citability, while an AI agent in the background points back into the system. The image explains that GEO Optimizer treats AI visibility as an inspectable workflow, not a vague marketing metric.
GEO Optimizer frames AI search as infrastructure you can audit, not just a leaderboard you can watch.
Key Takeaways

AI Search Is Not a Ranking Problem

Traditional SEO asks where a page appears in a results list. GEO Optimizer starts from a different question: will an LLM parse this page, trust it, and cite it inside an answer? That shift matters because the surface area is no longer a SERP. It is an AI system with its own retrieval, summarization, and citation habits.

The first open-source GEO audit and optimization engine. Based on the Princeton KDD 2024 research paper + AutoGEO ICLR 2026 methods.

Auriti Labs (GitHub Profile), Project Maintainer · Auriti-Labs/geo-optimizer-skill: GEO toolkit

That framing is the project’s real contribution. It does not treat AI visibility as a vague brand problem. It treats it as an engineering problem with inputs, checks, weighted scores, and outputs you can act on.

What GEO Optimizer Actually Does

The repository packages research on generative engine optimization into a practical audit stack. It checks whether a site is legible to machine readers across robots.txt, llms.txt, structured data, meta tags, content depth, AI-specific signals, discovery surfaces, and brand coherence.

The pipeline turns a page into a score, then into concrete fixes, with the same engine available to humans and agents.

A close-up of a scoring console with eight dials, some high and some low. One dial for schema richness is being adjusted by hand while a warning light for prompt injection sits nearby. The image explains that GEO Optimizer does not just check for the presence of signals. It grades their quality and flags risky content too.
The scoring engine is built to measure usefulness, not just presence.

The Scoring Engine Turns AI Readiness Into Numbers

The heart of the system lives in scoring.py. Its model breaks AI readiness into weighted categories, which is important because a page can be partially compliant and still be weak at being cited.

That is where schema richness becomes a useful distinction. A site that merely has JSON-LD is not the same as a site that expresses entities, relationships, and context in ways an LLM can actually use. GEO Optimizer scores the second case higher because it is closer to machine-readable meaning, not just markup presence.

# Conceptual shape of the scoring model
score = {
    "robots": ..., 
    "llms": ..., 
    "schema": {
        "presence": ..., 
        "richness": ...
    },
    "meta": ..., 
    "content": ..., 
    "signals": ..., 
    "ai_discovery": ..., 
    "brand_entity": ...
}

final_score = weighted_sum(score)