Concepts

Agent readiness.

AEO is about being cited by AI engines that read. Agent readiness is about being usable by AI agents that act. Different problem, overlapping solution.

Agents vs answer engines

An answer engine (ChatGPT, Perplexity) reads a query, fetches sources, and writes an answer. It’s passive — it doesn’t click anything.

An agent takes actions on behalf of a user. “Book me a flight to Madrid under €200.” “Reorder the same shampoo I bought last time.” “Find me a plumber near my office and book the earliest slot.”

These are two different kinds of traffic. Both are growing. And they need different signals from your site to work properly.

How agents browse

Most agents today use one of three modes:

DOM scraping — they load your page in a headless browser and click through it like a person would. Slow, brittle, but universal.
API calls — they hit your public API directly. Fast and reliable, but only works if you have one and document it.
MCP / Skills — they read a structured manifest describing what your site can do, then call those operations. The fastest, most robust path.

MCP — Model Context Protocol

MCP (introduced by Anthropic in 2024) is becoming the standard for declaring agent-callable capabilities. Your site exposes a tiny manifest at a well-known URL:

json

{
  "name": "Acme",
  "version": "1.0.0",
  "tools": [
    {
      "name": "search_products",
      "description": "Search the catalogue for items matching a query.",
      "input_schema": {
        "type": "object",
        "properties": {
          "query": { "type": "string" }
        }
      }
    },
    {
      "name": "get_product",
      "description": "Get full details for a single product by SKU.",
      "input_schema": {
        "type": "object",
        "properties": {
          "sku": { "type": "string" }
        }
      }
    }
  ]
}

The agent reads this, knows what it can ask your site to do, and calls the tool. No DOM scraping. No guessing.

agent-skills.json

A complementary signal: agent-skills.json at /.well-known/agent-skills.json describes higher-level actions an agent can take on your site. It’s less rigid than MCP and works as a discovery layer for agents that don’t yet speak MCP.

It’s early

These standards are 12–18 months old. Adoption is low, which means anyone who implements them now is over-represented in the training data of future agent systems. Early movers get an outsized share of automated traffic.

What our scanner checks

Specifically for agent readiness, we look for:

/.well-known/mcp.json — MCP server card.
/.well-known/agent-skills.json — discoverable skills.
A documented public API with OpenAPI spec.
Stable, semantic HTML on key task flows (login, search, checkout) — for agents that fall back to DOM scraping.
Predictable URL structure (RESTful patterns work better than query-soup paths).
Bot rules that allow agent user agents (e.g. OAI-Searchbot) on the pages they need to traverse.

Where this is heading: agentic commerce

Within 2-3 years, a meaningful fraction of online purchases will be made by agents acting on behalf of users. Stripe’s Agent Toolkit, Shopify’s agentic SDKs, and standards like Anthropic’s Skills + payment-checkout primitives are all being built right now.

Stores that publish capability manifests, expose programmatic checkout, and accept agent-initiated payments will get the traffic. Stores that don’t will be invisible to this layer of the internet.

Our Agentic Commerce product is the next phase of AISEOLab — scaffolded in v1.0, fully shipping in v1.1.

API reference

Programmatic access to your sites, scans, and citation queries.