The Product Data AI Shopping Assistants Actually Need
A shopping assistant does not browse your page. It tries to fill in a set of attributes so it can compare your product with others and answer a specific question. Whatever it cannot fill, it cannot use.
How an assistant reads a product
The question a shopper asks is usually narrower than your page: does this fit a size 42 wide foot, is it dishwasher safe, will it work with my model from 2019. The assistant resolves that into attributes, looks for values, and answers only when it finds them stated. Long persuasive copy does not help; a stated value does.
Attribute completeness is category-specific
There is no universal field list. Useful data means the attributes that decide a purchase in that category:
- Apparel — size system, fit, materials, care, measurements per size.
- Footwear — sizing system and conversion, width, upper and sole material, closure.
- Electronics — model, compatibility, power, connectivity, what is in the box.
- Home — dimensions with units, weight capacity, material, assembly.
- Consumables — volume or weight, ingredients, usage, storage, shelf life.
Identifiers make comparison possible
Brand, model and a valid GTIN let an assistant recognise that your product and a competitor's listing are the same thing. Without them, your product is an unmatched string and drops out of comparisons even when its data is good.
Structured data and feeds
Structured data on the page and a merchant feed serve the same purpose from different directions: they state facts in a form nothing has to infer. The rule that matters more than either format is that they must agree with the visible page. A feed price that differs from the page price makes both untrustworthy.
Language, country and currency are separate
Content language, target country, currency and market are four different decisions. A page written in English is not automatically priced in dollars or shipped to the United States. State the currency in ISO form, state the country you actually sell to, and never infer one from the other — assistants that guess get it wrong in the customer's favour, and you carry the complaint.
Gaps that block a recommendation
- No size or dimension data, so the product cannot be matched to a requirement.
- No compatibility statement in categories where compatibility is the question.
- Availability that is implied rather than stated.
- Claims with no supporting spec, which a careful system will not repeat.
Check your product data
See which of these attributes your product page actually states.