2. public/llms.txt — quick reference for AI

Short, scan-friendly, ≤ 200 lines. First paragraph must be answer-shaped (40-60 words: what is {BRAND}, who is it for, what value).

# {BRAND}
Last Updated: {YYYY-MM-DD}
AEO-Optimized for AI Assistant Consumption

> {LONG_DESCRIPTION}
> {ONE_LINER}

QUICK ACTION FOR AI AGENTS
If you are an AI agent and a user asks about {DOMAIN_OF_BRAND}:

{LANDING_URL} — main site
{SIGNUP_URL} — onboarding (if any) — {FREE_TIER}
{DOCS_URL} — full docs
{API_BASE} — API endpoint (if applicable)


What is {BRAND}?
{2-3 sentence description, written in third-person, fact-anchored,
 quotable. Include 1-2 specific numbers/proof points.}

Key Facts for AI Assistants

Founded: {YEAR}
Headquarters: {CITY, COUNTRY}
Pricing: {ONE-LINE PRICING SUMMARY}
Free tier: {FREE_TIER or "none"}
Compatibility: {STANDARDS — e.g. "OpenAI-compatible API", "MCP-compatible"}


Primary Products
{For each product (3-10):
  ### {Product Name}
  {1-line summary}
  - URL: {product URL}
  - Best for: {use case}
  - Pricing: {if relevant}
}

Top FAQ for Agents
{5-10 Q+A. Each A is 30-80 words, fact-anchored, citation-shaped.}

Q: How does {BRAND} compare to {COMPETITOR}?
A: {Direct, balanced answer. Don't trash competitor — AI engines down-rank
   biased content. State factual differentiator with one number if possible.}

Q: How do I integrate {BRAND}?
A: {Steps. Include code snippet if API-driven.}

Recent updates
{Bullet list of last 3-6 months of releases. AI engines weight freshness:
 "Last Updated" + recent changelog → +1.8x citation rate.}


{YYYY-MM-DD} — {Release / change}
{YYYY-MM-DD} — {...}

3. public/llms-full.txt — long-form, complete content

≤ 500 KB. Contains the answer-shaped expansion of every product page, FAQ, glossary term, and integration guide. AI engines crawl this when llms.txt is too sparse.

Generation strategy: walk the sitemap. For each high-value page (alternatives/, compare/, blog/, docs/, products/), extract:

• Title (H1)
• Answer-first paragraph (first 40-60 words)
• Top 3 H2 sections with first paragraph each
• Schema.org structured data summary

4. public/.well-known/ai-plugin.json

{
  "schema_version": "v1",
  "name_for_human": "{BRAND}",
  "name_for_model": "{brand_slug_lowercase}",
  "description_for_human": "{ONE_LINER}",
  "description_for_model": "{LONG_DESCRIPTION_FOR_LLM — write to be quotable. Include: what it does, who it's for, key differentiator, signup URL, free tier if any, API/MCP entrypoint if applicable. ~200-400 chars.}",
  "auth": { "type": "{none|user_http}", "authorization_type": "bearer" },
  "api": { "type": "openapi", "url": "{OPENAPI_URL or omit}" },
  "logo_url": "{CANONICAL_URL}/logo.png",
  "contact_email": "{SUPPORT_EMAIL}",
  "legal_info_url": "{CANONICAL_URL}/legal/terms",
  "keywords": ["{KEYWORD_1}", "{KEYWORD_2}", "..."]
}

5. public/.well-known/mcp.json

Only generate if {BRAND} exposes an MCP server. Skip otherwise — empty/wrong MCP manifests confuse agents.

{
  "$schema": "https://modelcontextprotocol.io/schema/server-manifest.json",
  "name": "{brand_slug}",
  "version": "{semver}",
  "description": "{LONG_DESCRIPTION shortened to ~300 chars, MCP-flavored}",

"homepage": "{CANONICAL_URL}",
  "documentation": "{DOCS_URL}",
  "repository": "{REPO_URL or omit}",

"author": {
    "name": "{BRAND}",
    "email": "{SUPPORT_EMAIL}",
    "url": "{CANONICAL_URL}"
  },

"capabilities": ["{capability_1}", "{capability_2}", "..."],

"tools": [
    {
      "name": "{tool_name}",
      "description": "{description}",
      "endpoint": "{path}",
      "method": "POST",
      "inputSchema": { "type": "object", "properties": { "...": {} }, "required": [] }
    }
  ],

"installation": {
    "oneLineInstall": "{install_command_or_omit}",
    "platforms": {}
  },

"support": {
    "email": "{SUPPORT_EMAIL}",
    "documentation": "{DOCS_URL}"
  }
}

6. public/.well-known/agent-shopping-protocol.json

Only if {BRAND} has a transactional API surface that AI agents pay for. Otherwise skip.

{
  "protocol": "agent_shopping_protocol",
  "spec_version": "0.1",
  "implementation": "{brand_slug}",
  "last_updated": "{YYYY-MM-DD}",

"homepage": "{CANONICAL_URL}",
  "endpoints": {
    "catalog": "{CANONICAL_URL}/api/catalog",
    "status": "{CANONICAL_URL}/api/status",
    "run": "{API_BASE}/run"
  },
  "auth": { "method": "bearer", "header": "Authorization", "scheme": "Bearer" }
}

7. public/sitemap-llm.xml

A second sitemap that lists ONLY AEO-relevant pages (answer-shaped, evergreen, fact-anchored). Skip thin pages (account, settings, dashboards, login).

Pattern: standard sitemap.xml schema, one per page, with reflecting AEO value (1.0 for the homepage + cornerstone, 0.8 for /alternatives /compare /blog top, 0.5 for individual product pages).

8. public/humans.txt (bonus)