LLM Readiness Audit

LLM & Agent Readiness Audit

This skill performs a marketer-focused audit of a website to determine how well it is optimized for the emerging "Agentic Web" (as highlighted at Google I/O 2026). Traditional SEO optimizes for Google Search; this audit optimizes for AI agents (Claude, Gemini, ChatGPT).

Goal

Measure the website's "AI Visibility Score" by checking for agent-specific standards and semantic clarity, providing diagnostic reasoning and actionable recommendations that marketers can pass to their engineering or CMS teams.

Audit Workflow

Step 1: Context Curation Check (llms.txt)

AI models use RAG (Retrieval-Augmented Generation) and web scraping to understand brands. An llms.txt file serves as a curated, high-fidelity index specifically designed for these models.

• Action: Attempt to fetch /llms.txt at the root of the provided URL.
• Evaluation: Does it exist? If so, is it formatted well?
• Marketer Context: If missing, LLMs have to guess what content is most important, which can lead to hallucinations about your brand's pricing, features, or messaging.

Step 2: Agent Interactivity Check (WebMCP)

WebMCP is the new open standard that allows websites to expose interactive tools (like "search inventory" or "book a demo") natively to AI agents.

• Action: Attempt to fetch /.well-known/mcp.
• Evaluation: Does the endpoint exist and return a valid manifest?
• Marketer Context: Without this, AI agents cannot natively take actions on your site. Users engaging with AI will have to leave their workflow and navigate your site manually, adding friction to the conversion funnel.

Step 3: Semantic Clarity Check

When agent-specific files are missing, AI relies entirely on crawling the HTML.

• Action: Analyze the homepage source for semantic HTML (e.g., <article>, <nav>, <main>) and Schema.org JSON-LD.
• Evaluation: Is the structure clean and easily parsable by an automated crawler? Does it avoid excessive client-side rendering blockages?
• Marketer Context: Clean semantics ensure that even basic AI scrapers accurately absorb the brand message.

Scoring & Report Generation

Calculate the AI Visibility Score (0-100):

• llms.txt Present: +40 pts
• WebMCP Endpoint Present: +30 pts
• Strong Semantic HTML/JSON-LD: +30 pts

Output the final report using the following markdown structure:

# 🤖 LLM & Agent Readiness Audit for [URL]

**AI Visibility Score:** [Score]/100
*This score reflects how easily AI agents (like Claude and Gemini) can understand your brand and interact with your website.*

---

## 1. Context Curation (`llms.txt`)
- **Status:** [Found / Not Found]
- **Why it matters:** AI models increasingly rely on Retrieval-Augmented Generation (RAG) rather than just crawling HTML. Without an `llms.txt` file, they have to guess what content is most important, which can lead to hallucinations about your brand's pricing, features, or messaging.
- **Recommendation for your Web Team:** [Provide a simple recommendation, e.g., "Create a markdown file at `/llms.txt` containing a 2-paragraph summary of your value prop and links to your 3 most important pages."]

## 2. Agent Interactivity (WebMCP)
- **Status:** [Found / Not Found]
- **Why it matters:** WebMCP is the new open standard from Google I/O 2026. It allows AI to natively perform actions on your site (like searching a catalog or booking a demo). If absent, you lose potential conversions from users who prefer AI-driven workflows and rely entirely on manual navigation.
- **Recommendation for your Web Team:** [Provide a basic suggestion on what tools they might expose, e.g., "Explore implementing a WebMCP endpoint at `/.well-known/mcp` to expose your product catalog directly to AI agents."]

## 3. Semantic Clarity
- **Status:** [Good / Needs Improvement]
- **Why it matters:** Clean semantic code is the critical fallback for basic AI scrapers. If your site relies heavily on complex JavaScript without semantic tags or Schema.org JSON-LD, automated systems will struggle to summarize your brand accurately.
- **Recommendation for your Web Team:** [Specific fix based on the audit, e.g., "Ensure core content is available in the initial HTML load and implement basic Schema.org JSON-LD for your primary entity."]

---
**Next Steps:** Would you like me to generate a draft of your `llms.txt` file or outline the specific JSON schemas your WebMCP server should expose? Let me know!