Product feed vs schema vs product-page content: where should AI-ready enrichment live?

Feeds are the most direct way to send structured product information into many commerce surfaces. They are particularly important where a platform explicitly accepts merchant product data, such as:

• Google Merchant Center;
• OpenAI's product feed, for approved partners;
• marketplaces;
• affiliate and commerce networks;
• shopping platforms and feed-management systems.

Feeds are strong for facts that need to stay synchronised: identifiers, price, availability, variants, shipping and product attributes. They can also carry the newer conversational and AI attributes where the destination supports them.

A feed also distributes exactly what the record holds. SO-ME, the consumer shopping platform Sighted powers, receives feeds from ASOS, REVOLVE, Vinted, M&S, New Look, Debenhams and EGO, and the same kind of product arrives with a different set of attributes and under a different category from one retailer to the next. A feed can only carry what the catalogue behind it holds.

A feed is not a replacement for a useful product page. Product pages provide richer context, supporting evidence, images, guides, FAQs, reviews, compatibility information and product-specific detail. They also remain crawlable sources for search and AI systems, and for ChatGPT the crawl is the main one: most of its product cards come from crawled pages rather than feeds, as the troubleshooting guide sets out.

Google's guidance on its AI features is clear that normal search fundamentals remain relevant: crawlability, useful text, internal linking, high-quality images and video, and accurate Merchant Center information still matter.

Product structured data helps search systems interpret details on a product page and can make pages eligible for relevant search features. But schema should not be sold as a standalone “AI visibility layer”. Google says no special schema is required to appear in its AI features, and asks for structured data that matches the visible page.

The rule is visible, accurate product content first, and a structured representation of that content second. Do not use structured data to claim facts that the page or the underlying product source cannot support. In Sighted's pilot with ChristianShirts.Net, the work was structured content and schema improvements, and AI citation coverage went from 0% to 27%.

For most retailers, the best approach is for Sighted to enrich the existing catalogue or product source rather than create a disconnected shadow catalogue. A sensible architecture runs from the retailer's catalogue, through Sighted's enrichment, to approved enriched fields, and out to each channel.

Those outputs may include product-page content on Shopify, product structured data, a Merchant Center primary or supplemental feed, an ACP or partner feed, and other supported destinations. The exact route depends on your stack. What matters is that the enriched information stays governed and can be updated when source facts change.

Avoid:

• publishing enrichment only inside Sighted, where external shopping systems cannot reach it;
• putting every generated attribute into visible product-page copy regardless of usefulness;
• stuffing schema with claims that are absent or unsupported on the page;
• overwriting retailer source data without governance;
• duplicating the same text indiscriminately across every product and channel.

1. Keep one trusted product source and enrich it, rather than building a disconnected shadow catalogue.
2. Publish channel-specific attributes into the feeds that accept them: Merchant Center, OpenAI's product feed for approved partners, marketplaces and networks.
3. Expose accurate product facts on crawlable product pages, with the context, guides, FAQs and compatibility detail a feed cannot carry.
4. Add structured data that describes the visible page, and never claims a fact the page or the product source cannot support.
5. Route approved enriched fields out to each channel under governance, and update them when source facts change.

Sighted adds no second place where you have to maintain product data. It is the layer that analyses the catalogue against real shopping intent, creates and structures validated enrichment, sends approved enrichment back into your existing destinations, and measures the resulting change.

It enriches the catalogue at scale, the technology behind SO-ME, and publishes to all three places this guide separates: approved values go into the product on Shopify, where they feed the page's structured data, the Merchant Center feed is built from the same records, and a store on another platform works from a CSV or feed URL. The free AI visibility audit reads up to 50 of your product pages and their markup first, so you know which of the three is the weak one.