Categorize & Classify Products Automatically with AI

Why categorization is harder than it looks

Assigning a product to the right category sounds trivial until you do it across a real multi-supplier catalog. Supplier A calls it "Herrenschuhe," supplier B "Men's Footwear," your shop calls it "Shoes → Men → Sneakers," and a marketplace wants an eClass code. None of those line up automatically. So someone sorts products into the tree by hand — and re-sorts them every time the tree changes or a new channel demands its own taxonomy. On a large catalog that's a permanent, low-value chore.

It's also high-stakes: miscategorized products don't show up in the right filters, get buried in navigation, and fail marketplace validation.

How AI assigns the right category

Rather than matching a keyword, AI categorization works from the whole product:

• It reads the attributes and description — and can use the product image — to understand what the product actually is.
• It places it at the right node in your tree, even when the supplier's category name is different or missing.
• It handles ambiguous cases a rule can't — a "3-in-1 jacket" that a keyword rule would drop in the wrong place.

Because it reasons from content, it degrades gracefully instead of failing hard when the data isn't perfectly labelled.

Your tree — and industry standards (ETIM, eClass)

Categorization isn't only about your own navigation. Selling into B2B catalogs or certain marketplaces means classifying products against a formal standard — ETIM, eClass, or a channel's own product taxonomy. AI can assign both from the same product data: your internal tree for the shop, the standard for the partners that require it. If those standards are new to you, start with GDSN, ETIM & eClass explained.

Doing it with a script — and where it stalls

A keyword or rule-based mapping handles the clean cases and is worth having for those. An LLM script goes further, classifying from the description. Both hit the same walls: products whose data doesn't contain the tell-tale keyword, a tree that gets restructured (now everything needs re-checking), multiple target taxonomies to maintain in parallel, and no review step to catch the confident-but-wrong assignments. Keeping that logic current across suppliers and channels is ongoing engineering.

How Productbay does it

In Productbay, categorization is part of AI Autofill: filter to the products you want, run it, and the AI assigns categories from attributes, descriptions and images — to your tree and, where needed, to a standard. Assignments land in the review queue so you approve in bulk and fix the uncertain ones. When your tree changes or you add a marketplace taxonomy, you re-run it on the affected products instead of re-sorting by hand. It shares the same data and review flow as enrichment and channel export, so a product is categorized once and stays consistent everywhere it's published.

This table was compiled from publicly available information. We aimed to bring transparency to the market — details may change over time. When in doubt: check both providers yourself and decide based on your own evaluation.

Categorization is one step in the flow; see the whole workflow in AI for product data maintenance, or how attribute mapping feeds it in bulk product data mapping.

Frequently Asked Questions