I build selling agents for webshops, and the first thing I check in a new store is the product data. Most skincare product pages I open describe a serum in marketing language and leave out the facts an AI agent needs to match for a recommendation: skin type, main active ingredient and its percentage, texture, fragrance, and also what the product should not be combined with.
Product data enrichment got fresh attention on 29 October 2026, when Coveo announced it had acquired Anglera, a company that uses AI to fill in missing product attributes. In the announcement, Anglera reports a 30-day test on about 300 product variants at one apparel retailer, with 120% more organic landing-page traffic and 270% more AI-referred visits for the enriched products. That is one vendor, one retailer and one category, and the release itself says results may vary. I read those numbers as a good reason to run your own test. They do not tell you what a skincare shop will see.
Most of this test plan is something a WooCommerce merchant can do without buying anything. Here is how we approach it.
It means writing down the facts a shopper would ask about, in fields a machine can read. A description tells a person a story. Attributes let an agent filter consistently. If a shopper asks ChatGPT for a fragrance-free niacinamide serum for oily, acne-prone skin, the agent needs fragrance, active ingredient and skin type as clear values. When those facts live only inside a paragraph of brand copy, or only on a photo of the label, the agent has to guess, or it moves on to a store where it does not have to.
We see the same thing on the merchant side. Our agent asks shoppers about skin type and concerns before it recommends anything, and it reads the same attributes to answer. When the attributes are clean, the questions get shorter and the recommendations get more accurate, whether the buyer is a person on your site or an AI agent shopping for them.
Two public specifications show what the large platforms ask for.
OpenAI's product feed spec for ChatGPT requires nine fields: item_id, title, description, url, brand, seller_name, image_url, availability and price. It asks for concise titles (up to 150 characters) and plain-text descriptions (up to 5,000 characters), and it calls the description a "factual product description". Optional fields include material, size, width / length, color, product_category, gtin code, review_count and star_rating. The spec defines no general field for custom attributes, and it says question-and-answer lists are not part of this discovery feed.
Google Merchant Center has a product_highlight attribute for short highlights of 1 to 150 characters each, with 4 to 6 recommended per product. Google says highlights can help products appear in AI-driven surfaces such as AI Mode in Google Search. Examples are characteristics like "fragrance-free" or platform support if you sell electronics.
For a skincare shop, the facts that decide a match (skin type, key active, fragrance) have no dedicated field in the ChatGPT feed. So they need to be written clearly in the title, the first lines of the description and the highlights, and stored as attributes on your own site, where your search and your assistant can use them.
This is the list I start with when we set up a store. Not every product needs every field, but every product in a category should use the same ones.
WooCommerce has two kinds of attributes. Global attributes live under Products > Attributes and, in the WooCommerce documentation, "are available to be used on any product". Custom attributes are added on a single product and "are not available for global use".
Use global attributes. They keep values consistent across the catalog, so you have one spelling of "fragrance-free" instead of five, and WooCommerce can use them in the Product Filter block, which helps the shoppers on your own site as well.
Then carry the same facts into the text fields. Put the main active and the skin type in the title when it reads naturally. Open the description with two or three plain sentences about who the product is for and what is in it, and put the brand story after that. If you send a feed to Google, map four to six of these attributes into product_highlight.
Run a smaller and more careful version of the test in the Anglera announcement. Pick one category, for example serums. Split it into two groups that had similar traffic over the past months. Enrich one group and leave the other alone. Wait at least 30 days, then compare organic sessions, visits referred by AI assistants, and orders between the two groups.
Expect small numbers. A skincare shop with a few hundred products will rarely get a clean answer in one month. The test is still useful, because it shows you a large effect if there is one, and it gives you your own baseline to hold any vendor's claims against.
Want to see how AI agents read your store today? We built a free snapshot for WooCommerce shops.
What is product data enrichment?
Product data enrichment means adding structured, factual details to each product, such as skin type, key active ingredients and texture, so search engines, AI shopping agents and your own site search can match products to what shoppers ask more accurately.
Does ChatGPT read WooCommerce product attributes?
ChatGPT shopping works from a the OpenAI product feed with nine required fields, including title and description. The OpenAI spec has no general field for custom attributes, so skincare facts such as skin type and key actives should also appear in plain text in the title and description.
Should I use global or custom attributes in WooCommerce?
Use global attributes, created under Products > Attributes. You can reuse them on every product, they keep values consistent, and you can use them in product filters. A custom attribute exists on one product only.
Do I need a paid enrichment tool?
Not to start. A small skincare catalog can be enriched by hand from the product labels. Run your own test on one category first, then decide whether a tool would save enough time to be worth the price.