ShoppingAttribotShoppingAttribot“Does it fit in the dishwasher?” “What else should I buy with it?” “Is it suitable for allergy sufferers?” This is how people shop today — the answer is built from whatever the system knows about the product from your feed. If that information isn't there, someone else's product gets picked instead.
No installation, no changes to your store — just an XML feed.
Google introduced a separate set of attributes for AI surfaces — the exact ones answers get built from. Right now, hardly anyone has filled them in, so the gap between a feed that answers and one that just describes is bigger than it will ever be again. Once everyone fills in this layer, it stops being an advantage and becomes a requirement.
Five attributes from Merchant Center's conversational set. Three are always generated; two variant ones are filled in only for feeds carrying an item_group_id — that's noted on the card.
Pre-purchase questions with a factual answer — volume, material, compatibility, care. Built from the product's title, description, and specs.
Links to accessories, required parts, often-bought-together products, and alternatives from other brands. Only ever points to products actually in your feed.
The product's percentile within your own assortment. Computed from your GA4 data — revenue, units, or clicks.
A shared name for all variants of one model.
Lists what makes the variant different — color, size, material.
Four steps. Between uploading the feed and downloading the file, there's nothing to configure in your store.
An XML feed from Merchant Center. The file is parsed directly in your browser — only the selected part is sent to the server for processing.
The whole catalog, specific IDs, or a manual pick. You don't have to start with everything at once.
An export of revenue, units, or clicks per product. Without it, popularity isn't computed — the other attributes are generated either way.
TSV for Merchant Center or CSV for Mergado. Upload it as a supplemental data source — it matches by id and doesn't overwrite the original feed.
A coffee machine with an ordinary description, the kind almost every store has in its feed.
A descaler, a tamper, and a milk pitcher — all from the same catalog.
Percentile by revenue within your assortment.
id popularity_rank question_and_answer related_product
SKU-1042 96.4 "Does the machine fit under a kitchen wall cabinet?":"The unit is 38 cm tall, 32 cm wide, and 42 cm deep — it fits under a standard cabinet mounted at 45 cm.","Can it froth plant-based milk?":"Yes. The steam wand is adjustable and designed for both dairy and plant-based milk." required_part:id:SKU-4471* Anonymized example — the product and identifiers are fictional; the data structure matches the real output.
The questions we get most often — the ones we'd ask too.
No. The attributes go into a supplemental data source that's matched to your primary feed only by id. It doesn't overwrite titles, descriptions, or prices, and doesn't change approval status.
The model can only draw on the product's own title, description, and specs. If nothing factual follows from them, or the description doesn't match the product, it leaves the field empty instead. An empty result is the intended outcome, not a bug.
You upload an XML feed and download a file. No installation, no changes to your store, nothing connects to your system.
You choose the output language. Questions and answers are generated in it regardless of what language the feed's descriptions are in.
Yes. You see an estimate before starting and the actual usage once it's done. You only pay for products that actually received content.
Details on how we handle data are in the How we handle data section.
Pick a handful of products, look at the result, and only then decide. It costs a few crowns and a few minutes — and you'll see exactly what you'd get for your entire assortment.
Upload the XML, download the file. Nothing connects to your store.
Attributes go alongside it, into a supplemental data source.
Only for products that actually received content.