Shopping agents read your catalog, not your homepage

A shopping agent doesn't look at your homepage hero or your brand video. It reads your catalog: titles, attributes, prices, stock, variants. It picks from that data, then hands the shopper to your store.

Two things that happened in 2026 make this concrete.

Shopify: catalog data converts twice as well

On Shopify's Q2 2026 earnings call, its president, Harley Finkelstein, said "AI searches powered by Catalog converted twice the rate of those using scraped data." Shopify Catalog is the structured product data Shopify makes available to AI channels.

The comparison is the useful part. Shoppers searching through AI converted at twice the rate when the agent worked from structured product data instead of whatever it could scrape from a page.

He also said 75% of AI-attributed purchases came from outside the top 100 categories. Agents match products to specific intent, so the long tail gets found, as long as its data says what it is.

OpenAI: discovery in chat, checkout on your site

In September 2025, OpenAI launched Instant Checkout, which let shoppers buy inside ChatGPT. On March 24, 2026, it changed course: "the initial version of Instant Checkout did not offer the level of flexibility that we aspire to provide, so we're allowing merchants to use their own checkout experiences while we focus our efforts on product discovery."

So for now, the assistant finds the product, and your store makes the sale.

Your data works at both ends

Put those together and catalog data does two jobs.

• Discovery. An agent can only choose what it can read and match. A product whose material, size, use case, and compatibility live only in a photo or a vague paragraph can't be matched to a shopper's requirements.
• The sale. The shopper lands on your product page to finish what the agent started. A price, stock level, or variant that disagrees with what the agent said is a reason to leave.

The costliest failures are mismatches: the feed says in stock and the page says sold out, or the agent quotes $109 and the page says $129. Each one costs trust at the exact moment the shopper was ready to buy.

Where to start

1. Fill the attributes shoppers filter by: material, size, fit, compatibility, audience. Put them in metafields so they're structured, not buried in prose.
2. Group variants properly, so one product in five colors isn't five weak listings.
3. Add identifiers (GTIN, MPN) where you have them, so an agent can tell your listing is the same product it has seen elsewhere.
4. Keep price, stock, and variants identical across the page, the structured data, and any feed.
5. State shipping and returns on the product page, because agents and shoppers both ask.

What you can't control

Which agents your shoppers use, how each one ranks candidates, and how fast the platforms change. Instant Checkout lasted about six months. The specifics will keep moving; the dependence on clean catalog data is the part that has held.