Can an AI agent shop your store?
Type in a store. Hunch runs the six read-only checks the merchant-readiness index runs every week on the largest US retailers, and tells you what a shopping agent can reach, what it cannot, and what to fix first.
Six plain GET requests under an honest user agent. Nothing is logged in, submitted, carted or stored beyond a daily count of checks. How it works.
Raw evidence (what each request returned)
The checks and weights come from the index's method. Score is out of 100: the homepage carries 35 because an agent that cannot read it cannot do anything else. Want the layer under this, from the mandate to settlement? The free course Anatomy of an Agentic Payment walks one transaction step by step.
What the six checks look at
The same probes, in the same order, as the weekly index. Only your own report is shown to you; nothing is added to the public data.
robots.txt
Is it published and readable, and what does it tell 14 named AI crawlers and agents: training bots, AI search, and agents acting for a live shopper.
The homepage
What a self-identified non-browser agent gets back: a page with links and a title, a bot challenge, or an empty shell that only fills in with JavaScript.
Structured data
JSON-LD on the homepage, and whether any of it is Product, Offer or ItemList, which is what lets an agent read price and stock without scraping.
llms.txt
A plain-text guide to the site for language models. Cheap to publish; 15 of the 92 largest US retailers had one in October 2026.
UCP profile
The Universal Commerce Protocol business profile at /.well-known/ucp, which tells an agent where checkout and policies live.
A2A agent card
An agent-to-agent card at /.well-known/agent-card.json. Rare so far: the index found its first in a large US retailer in October 2026.
The playground: a store built to be broken
A reference shop that is consistent everywhere an agent might look, ten traps built wrong on purpose with an answer key, a mock checkout that enforces a spending mandate, and a run record that shows what your agent actually did. Point an agent at it and compare its story with the record.
The playground and its report card are live. New traps are added from time to time: one email when they land, nothing else.