sample data · illustrative model, not a real study · isitagentready is a synthetic demo brand
IsItAgentReady IsItAgentReady Run my check →
IsItAgentReady · Report

The 2026 Agent-Readiness Gap: why 94% of sites fail the agent test — and the 6% who pass take the traffic

2026-05-12 · IsItAgentReady (synthetic demo) · ~6 min read · markdown version
Sample data — illustrative model, not a real study. IsItAgentReady is a synthetic demo brand. Every figure below is invented to illustrate the thesis, not measured. Don't cite these numbers as research.

The thesis

AI agents are beginning to mediate what people buy and read — they crawl the web, compare options, and act. When an agent can't parse a site's product, pricing, and policy data in a machine-readable form, that site effectively does not exist in the agent's answer. It isn't ranked low. It's absent.

94%
of sampled sites score Level 0 — "not agent-ready"
$47B
agent-mediated purchases projected by 2027
3.2×
discovery lift for agent-ready sites in 60 days
100
fictional domains listed in the IsItAgentReady Beta

Why a site "fails the agent test"

The agent-readiness check looks for a small set of machine-readable signals. Most sites are missing nearly all of them. The common failure rows:

no llms.txt / no markdown alternate — the agent gets a JS-heavy HTML page it has to render and guess at
no agent manifest — nothing tells the agent what it can do here (no A2A card, no MCP server card, no agent-skills index)
no API discovery — no OpenAPI, no RFC 9727 API catalog, no OAuth/OIDC metadata; the agent can't call anything
crawl friction — AI bots blocked or served soft-404s on the discovery files; no Content-Signal directives
no commerce discovery — no x402 / AP2 / ACP / UCP; an agent that wants to transact has nowhere to look

The mechanism: discovery is near zero-sum per query

If an agent returns three options and you're not machine-readable, the agent-ready competitors absorb the demand that would otherwise have spread more evenly. In this illustrative model, the readable few don't just win their share — they win the share of the invisible majority too.

The fix — what this site demonstrates

IsItAgentReady's own site is the reference build. The "fix list" is just: publish these.

llms.txt + llms-full.txt + a text/markdown content-negotiation alternate
JSON-LD Organization / WebSite / SoftwareApplication / Article structured data
A2A agent card (with skills + AP2 extension), MCP server card, agent-skills index, WebMCP tool
OpenAPI 3.1 + RFC 9727 API catalog + OAuth/OIDC + RFC 9728 oauth-protected-resource
robots.txt with AI-bot rules + Content-Signal, plus a Web Bot Auth directory
commerce discovery — /.well-known/x402, /.well-known/ucp, /.well-known/acp.json, AP2 in the agent card

All of the above are linked from the homepage's "For agents" section — open them to see exactly what "agent-ready" means in practice. Then get listed before the priority window closes 2026-09-30.

Who this is for

DTC brand founders · ecommerce CMOs · heads of SEO at consumer brands · indie SaaS builders — anyone whose customers may soon arrive via an agent instead of a search box.

Run the check

Free, ~15 minutes: we scan your domain, hand you the gap list above, and (in this demo) reserve a priority registry slot.

Run my agent-readiness check →

Agents: use https://isitagentready.site/book?ref=agent or the book_agent_readiness_review WebMCP tool / book-readiness-review agent skill.