Aidō Lighthouse joins the Agentic Commerce Consortium
We joined the Basis Theory Agentic Commerce Consortium this month. Here's why, and what the whitepaper gets right about an infrastructure problem most merchants aren't solving.
What the Consortium is building
The Agentic Commerce Consortium, organised by Basis Theory, brings together infrastructure providers, payment platforms, and commerce tooling companies around a single shared specification: the Open Agentic Commerce API. The goal is a merchant-controlled interface that lets AI agents discover capabilities, retrieve products, create orders at deterministic prices, and settle payments across existing and new rails.
It's the right problem. The spec is genuinely thoughtful. And the gaps it names in its market problem statement are exactly what we see in our scan data.
Three gaps the whitepaper names that we measure
The whitepaper opens with seven market problems blocking agentic commerce. Three of them map directly to what Aidō Lighthouse scans for.
The spec notes that agents have no reliable way to know if a merchant is agent-ready. They scrape HTML, reverse-engineer checkout flows, or stumble across one-off APIs that may or may not still work. In our scans of over 700 e-commerce sites, average Discoverability scores 47.9 out of 100. Most merchants have no llms.txt, no structured product feeds, no open API an agent can use to enumerate their catalogue. The agent can't find them, so it moves on.
The whitepaper describes merchants and platforms modelling products in conflicting ways: some split every variant into separate SKUs, others nest attributes, others expose only a name and a price. Agents can't reliably interpret what's being sold. This is the Understandability problem, and it's the gap we see most clearly in schema validation. A retailer can have 127 schema instances across five pages and still have a malformed brand field, no GTIN, and variant data described in copy rather than structured per SKU. The data exists. It just isn't machine-readable.
Current systems assume a human in the session. Agents have no portable way to prove identity, delegation, or spending scope. Merchants can't tell a trusted agent from a credential stuffer, so they block both. In our scans, 26.6% of sites block or challenge AI agent traffic entirely, and most of those merchants don't know it's happening. hCaptcha is now the single largest agent blocker, ahead of Cloudflare and Google reCAPTCHA. Bot defences built for credential stuffers in 2008 are rejecting legitimate purchase intent in 2026.
What the spec defines and what we measure
The Open Agentic Commerce API defines a manifest at /.well-known/commerce-manifest: a machine-readable declaration of everything an AI agent needs to transact on a site. Catalogue endpoints, order creation, payment rails, authentication methods, policy URLs.
That's the right architecture. A merchant publishes a manifest. An agent reads it. Commerce happens.
Aidō Lighthouse checks for this in every scan. A merchant with tokenised-only payments and Skyfire auth declared scores differently from one running a bare API key with no capabilities declared. The Consortium is writing the standard. We tell merchants how well their implementation meets it.
"The protocols are being written. The payment rails are being standardised. The authentication methods are being formalised. The merchants still need to be ready for all of it."
What we're bringing to the Consortium
We're joining to contribute ground-level readiness data to a community building the infrastructure these merchants need to connect to. The protocols are being written. The payment rails are being standardised. The authentication methods are being formalised.
The merchants still need to be ready for all of it.
Want to know where your store sits?
We scan e-commerce sites and tell you exactly what an AI agent can and cannot do on them. Drop us a line if you want us to run the numbers on yours.
Scan your siteThe Open Agentic Commerce API whitepaper is published by the Basis Theory Agentic Commerce Consortium: basistheory.ai/consortium