Categorizing Messy Supplier Feeds with AI in WooCommerce

One of the most frustrating aspects of importing dropshipping or supplier feeds is dealing with chaotic taxonomies. Your store might use a clean category structure like “Men > Apparel > T-Shirts”, but your supplier’s XML feed lists the item under a useless category like “Summer Promo / Overstock / Misc”.

Historically, fixing this required building massive, hardcoded Excel mapping tables or writing hundreds of “IF/THEN” PHP statements. Today, Large Language Models (LLMs) have completely revolutionized complex data sorting.

1. The Death of Manual Category Mapping

If you rely on exact-match rules to map supplier categories to your WooCommerce categories, your import will fail the moment the supplier introduces a new, unmapped label. This leaves your catalog filled with “Uncategorized” products that customers cannot find.

2. Using LLMs as a Data Router

Instead of relying on rigid rules, we can place an AI model (like OpenAI’s GPT-4o) between the incoming supplier feed and your database. During the import process, a custom PHP script fetches your store’s exact WooCommerce category tree.

The script then feeds the LLM the raw supplier product name, its description, and your strict category list, instructing the AI: “Review this product data and reply with ONLY the exact ID of the most relevant category from our provided list.”

3. Forcing Strict JSON Output

To ensure the AI doesn’t hallucinate or reply with conversational text, the PHP parser forces the LLM to output its decision in strict JSON format.

• The Input: Supplier Title: “Razer BlackWidow V3”. Supplier Category: “Misc Hardware”.
• The AI Processing: The LLM recognizes it as a mechanical keyboard.
• The Output: {"category_id": 145} (Mapping perfectly to your “Gaming Keyboards” taxonomy).

The PHP script intercepts this JSON, assigns the ID using wp_set_object_terms(), and inserts the product cleanly.