JD | building Cartable

Shopify × AI agents. Building Cartable (open-rubric store audits) + side projects in public. 52-store audit: 0 block AI bots; 81% miss return/shipping data.

[JD | building Cartable @JD_eth]

@vikrantshukla From our 51-store audit, it's not even a product field. It's policies: ~4 in 5 stores had no machine-readable shipping or return info, so an agent can't tell what the item really costs to get.

[JD | building Cartable @JD_eth]

@DarshanD3v @vercel @Shopify Exactly. On Shopify it's usually a few lines of shippingDetails and hasMerchantReturnPolicy in the product JSON-LD, and most themes just don't ship it. It's one of the first things the Fix Pack covers.

[JD | building Cartable @JD_eth]

@siddhibee Policies, by a wide margin: in our 51-store re-audit ~4 in 5 had no machine-readable shipping or return info. Your dead-button case is scarier though, since nothing on the store side logs it, so it never gets fixed.

[JD | building Cartable @JD_eth]

@theashbyte Exactly. Agents read the product page as well as the catalog, so hiding it in one place doesn't fix it.

[JD | building Cartable @JD_eth]

@DarshanD3v @vercel @Shopify Policy data, by a long way. In our 51-store re-audit ~4 in 5 miss shipping/return markup vs about half with og gaps. Policies usually sit behind a footer link or app widget, so an agent on the product page can't tell shipping cost or whether returns are allowed.

[JD | building Cartable @JD_eth]

@TaylorSicard The checkout layer keeps changing hands (Instant Checkout, now Shop MCP, Google's agentic checkout), but every one of them reads the same product data. Stores that make price, variants and return terms machine-readable win no matter which one sticks.

[JD | building Cartable @JD_eth]

@DarshanD3v @vercel @Shopify That's basically how our audit works: raw HTML only, no JS, then we check that the name, price and add-to-cart form are actually there. Most themes are fine on name and price. The trouble is the meta and policy data.

[JD | building Cartable @JD_eth]

@reversible_labs It does. We check that each product exposes its variants in /products/<handle>.js and in Product JSON-LD with one Offer per variant (price, stock, SKU). Unnamed option values are a common miss, agreed.

[JD | building Cartable @JD_eth]

We re-audited 51 well-known Shopify stores on 20 Oct to see if an AI shopping agent can find, read and add a product to cart.

Median score: 86/100. 80% (41 of 51) don't mark up return or shipping terms.

Full ranking + top gaps per store: https://t.co/ZGCnbjrwOV

[JD | building Cartable @JD_eth]

@KOF97AI Yes, they're separate checks. One tests a real add-to-cart call (no checkout). Another looks for shipping and refund policy pages plus MerchantReturnPolicy and shippingDetails markup. Most stores pass the first and fail the second, which is where the 80% comes from.

[JD | building Cartable @JD_eth]

@DarshanD3v @vercel @Shopify Seeing this in our audits too. 28 of the 52 Shopify stores we checked had no og:price:amount tag, so an agent that doesn't run JavaScript has to guess the price. Server-rendered Product JSON-LD with price, stock and variants is the cheapest fix.

[JD | building Cartable @JD_eth]

@frombailee @vercel @Shopify Honestly that's the unsolved part. Shopify's checkout MCP now asks for buyer consent, but the record of what was approved sits with the agent platform, not the store. Until that's standard, the best defence is a return and shipping policy the agent can read before buying.

[JD | building Cartable @JD_eth]

@ethanwalkerman Fair point. Agents skip the upsell layer, so margin per order can drop. The stores that win make what the agent weighs visible: shipping cost, return window, bundle pricing, so it's not just sticker price. Readable is step one. Pricing for agent traffic is step two.

[JD | building Cartable @JD_eth]

@whoinvitedethan Getting blocked is the obvious way to lose the sale. The quieter way is a store that lets the agent in but leaves out shipping cost or the return window, so the agent picks a competitor that spells it out. Worth checking both.

[JD | building Cartable @JD_eth]

@frombailee @vercel @Shopify Returns are where it gets messy. When we audited 52 Shopify stores, 81% didn't publish return or shipping info in a form an agent can read, so the agent can't even check the policy before it buys. Fixing that first would head off a lot of these disputes.

[JD | building Cartable @JD_eth]

@nielskaspers This is the part most stores miss. When we audit Shopify stores, price is usually the gap: 28 of the 52 we checked don't publish og:price:amount, so an agent has to guess. Your four points make a good checklist.

[JD | building Cartable @JD_eth]

@1RustyMac @elonmusk Agreed, that's a clean split. Cartable checks whether an agent can shop the store, and agent-welcome says whether it may. I'd like to add a check for an agent-welcome file to our audit so stores get credit for publishing one. Mind if I DM you to line up the format?