How I Built an AI Powered Bulk Product Importer for WooCommerce Using Google Drive

Adding a few products to WooCommerce manually is simple. Doing the same thing for a large product catalog is a completely different problem.

For this project, I was working with a jewelry store where product images were already organized in Google Drive. Each product needed a name, description, category, material information, multiple images and other product specific data before it could be added to WooCommerce.

Doing all of this manually for every SKU would take a lot of time and would also make the process repetitive.

Instead of building another basic CSV importer, I developed a custom WordPress plugin that combines Google Drive, AI image analysis and WooCommerce into one product import workflow.

The Problem

The product images followed an SKU based naming structure.

For example:

B087.jpg
B087a.jpg
B087b.jpg

These are not three different products.

They are three images belonging to the same product with the SKU B087.

The system therefore needed to understand the relationship between filenames before doing anything with AI or WooCommerce.

There were several other requirements as well.

The importer needed to generate useful product information from images, work with multiple images for each SKU, use the store's existing WooCommerce categories, handle materials and custom fields, create structured CSV data and finally import everything into WooCommerce.

This turned the task into more than a standard product importer.

The Workflow I Designed

I separated the process into several stages instead of asking AI to control the entire import.

The basic workflow became:

Google Drive → Image Discovery → SKU Grouping → AI Analysis → Structured Product Data → CSV → WooCommerce

This separation was important.

AI is useful for understanding what is visible in an image and generating descriptive content. It shouldn't be responsible for every business decision in an ecommerce catalog.

The plugin controls the workflow while AI handles the parts where visual interpretation is actually useful.

Grouping Google Drive Images by SKU

One of the first problems was determining which images belonged together.

Instead of treating every Drive image as a separate product, I added logic to recognize the base SKU.

For example:

E174.jpg becomes the primary product image.

E174a.jpg and E174b.jpg are detected as additional images belonging to E174.

Once the images are grouped, the plugin can process the complete image set as a single product.

This also means the WooCommerce product can receive its main image and gallery images automatically instead of requiring them to be assigned manually later.

Using AI Vision for Product Information

Once an SKU and its images are identified, AI vision can analyze the product.

For jewelry products, the visual information is useful for generating fields such as:

Product Name

A short product name based on the design and visual characteristics.

Description

A customer friendly description based on the visible shape, colors, materials and design.

Inspired By

A more creative description of the visual inspiration behind the piece.

Material

Material related information that can then be mapped into the store's product structure.

The important part was not simply asking AI to describe an image.

The response needed to follow a predictable structure so that the plugin could convert it into usable ecommerce data.

Working With Existing WooCommerce Categories

Category handling introduced another interesting problem.

Allowing AI to freely generate categories can quickly make a WooCommerce catalog messy.

You could end up with categories such as:

Earring
Earrings
Designer Earrings
Jewelry Earrings

even when the store already has a properly organized category structure.

I therefore designed the workflow around the store's existing WooCommerce categories.

The existing taxonomy acts as the source of truth. The product can be analyzed and matched with an appropriate existing category instead of creating unnecessary taxonomy terms.

This keeps AI generated information aligned with the actual store structure.

Generating Structured CSV Data

After processing a product, the information is converted into a consistent CSV structure.

The importer works with fields including:

SKU
Product Name
Price
Description
Images
Categories
Inspired by
Material

This intermediate CSV layer is useful because it gives me a structured checkpoint between AI generation and the final WooCommerce import.

The generated information can be reviewed before it becomes live product data.

Handling Custom Product Data

The store wasn't limited to standard WooCommerce fields.

Some product information also needed to work with custom fields and multi select values, including material information.

That meant the importer couldn't simply create a product title and description.

During import, the plugin also needed to map generated values into the correct WordPress, WooCommerce and custom field structures.

This is one of the reasons I chose a custom plugin instead of trying to connect several unrelated import tools together.

AI Shouldn't Guess Everything

One lesson from this project was deciding which information AI should and shouldn't generate.

A product description can reasonably be created from an image.

A price cannot.

Business critical information such as pricing should come from an authoritative source rather than being guessed from visual information.

I therefore treated AI as one component of the import system rather than the database itself.

The same principle applies to categories. AI can help determine the most appropriate category, but the store's existing taxonomy remains the source of truth.

Challenges During Development

Several issues only became obvious once I started testing the workflow with real products.

Multiple images needed to be grouped correctly instead of creating duplicate products.

Generated categories needed to match categories that actually existed in WooCommerce.

Material values needed to work with the site's field structure.

Pricing needed to remain separate from AI generated visual data.

Batch processing also had to work reliably when multiple SKUs were submitted rather than only succeeding with a single test product.

These are the details that turn an AI demo into something that can actually be used inside a WordPress ecommerce workflow.

Why I Used a Custom WordPress Plugin

It would have been possible to perform parts of this process with separate tools.

The problem is that the business logic was specific to this catalog.

The SKU naming convention, Google Drive image grouping, AI instructions, WooCommerce taxonomy, custom fields and import rules all needed to work together.

Building a custom plugin gave me control over the complete process without changing WooCommerce itself.

It also means the workflow can be improved as the catalog requirements change.

The Final Workflow

The final system reduces a repetitive product entry process into a much more manageable workflow.

Instead of manually opening every image, identifying the product, writing content, organizing multiple images and preparing WooCommerce data, the plugin handles much of that processing automatically.

The user can focus on reviewing the generated information rather than creating everything from scratch.

For me, the interesting part of this project wasn't simply adding AI to WordPress.

It was deciding where AI was useful, where normal application logic was better, and how both could work together inside a real WooCommerce workflow.

That's the approach I prefer when adding AI to ecommerce projects. AI handles interpretation and repetitive content work, while the application remains responsible for structure, validation and business rules.