Can AI clean a messy supplier file so my Shopify CSV import works?

The problem, briefly

Shopify's import is strict. The first line must be column headers, columns must be separated by commas, and the file must follow the product CSV layout. Variants (a size or color of the same product) must share a URL handle, which is the unique web-address name of the product. Supplier files, by contrast, come from Excel exports, price lists and PDFs turned into sheets. They use their own column names, put a size and color in one cell, and repeat the full title on every row. Fixing that by hand for 400 products is a long afternoon of copy and paste. For the problem itself, start with my guide to Shopify CSV import errors and how to fix them.

That reshaping is exactly what a language model does well. The parent topic, the common Shopify import errors and their fixes, is the checklist. This article is about letting AI do the repetitive part. If your real issue is turning many kinds of paperwork into data, my guide to AI document processing covers the wider idea.

Where AI genuinely helps with this specific problem

• Renaming and reordering columns. Map the supplier's "Item Name", "Cost" and "Qty" to Shopify's Title, Cost per item and Inventory quantity columns, and drop what you do not need.
• Making handles. Shopify says a valid URL handle can contain letters, dashes and numbers but no spaces. A model can turn "Black Sunglasses (XL)" into black-sunglasses safely and keep it unique.
• Splitting variants. A cell reading "Red / Large" can become Option1 value and Option2 value. For products with variants, Shopify's rule is that only the first row carries Title, Description, Vendor and Tags, so a model can blank those on the following rows.
• Cleaning text. Tags with symbols, prices with currency signs (Shopify wants the number only, like 9.99), and weights in pounds that must be grams (Shopify's example: 4 pounds is 1814).
• Writing a list of problem rows. Missing prices, duplicate SKUs and broken image links get flagged instead of silently guessed.

Each is a narrow transformation with rules written down by Shopify, so the model has a clear target.

A real example

There is no published case study of a store using AI to fix Shopify CSV imports with measured results. What exists is smaller but real: Shopify's own AI assistant, Sidekick, which its help center says can create content such as product descriptions and write code to summarize store data. That is about editing products already in your store, not repairing an incoming supplier file, so it is a near neighbor, not a match. Treat the method below as my practice rather than a documented result, and measure it on your own file.

How to do this yourself with Claude

You do not need any special tool. You need Shopify's rules, a small sample, and a strict prompt.

1. In Shopify, go to Products, then Import, and download the sample CSV from the help page. Use its header row as your target. Delete its two example products.
2. Take 20 rows of the supplier file, including one product with several variants and one with missing data. Paste them into Claude along with the Shopify header row.
3. Send this: "Convert these supplier rows into Shopify product CSV format using exactly this header row. Rules: Title is required. URL handle is lowercase, letters, numbers and dashes only, no spaces, and unique per product. For a product with variants, put Title, Description, Vendor and Tags only on the first row; later rows keep the handle and variant columns. Option1 name and Option1 value must both be filled for any row that has a SKU or weight. Price is a number only. Weight is in grams as a whole number. Do not invent any value. If a value is missing, leave the cell empty and list that row in a separate Problems list with the reason. Output the CSV, then the Problems list."
4. Read the output against the source for five random rows. Check the Problems list.
5. Import those 20 rows. Shopify's help page says a CSV import cannot be canceled once it starts and you cannot see a history of past imports, so a small test first is not optional. It also advises a backup of your product data before you begin.
6. If the 20 rows land correctly, run the rest in batches. Shopify says the product CSV cannot exceed 15 MB. Split larger files.

Two checks matter because they bite quietly. First, do not sort the finished file in Excel or Apple Numbers: Shopify warns that a CSV sorted in a spreadsheet editor may cut products from their image links, so images are lost. Second, if you import over existing products with "Overwrite products with matching handles", Shopify says a blank non-required column overwrites the stored value with blank. A tidy-looking file with an empty Vendor column will wipe your vendors. Also remember that changing Option values deletes existing variant IDs and creates new ones, which can break apps that depend on them.

A worked example with made-up figures: a supplier sends 300 rows for 120 products. After the first prompt, the model returns 296 usable rows and a Problems list of 4 rows with no price. You fix those four with the supplier and the import succeeds, instead of spending a day on 300 rows by hand.

Where a person still checks

Prices, costs and stock numbers: a wrong decimal is a wrong price on your store. Product categories and tags need judgment. And file format: Shopify wants the CSV saved in UTF-8 with LF-style linefeeds, so confirm the save settings of your spreadsheet program. The CSV format itself has rules a model can forget. The standard that describes CSV, RFC 4180, says a field containing a comma, a double quote or a line break must be wrapped in double quotes, and a quote inside a field is written as two quotes. A supplier title such as "Mug, large, blue" splits into three columns if the quotes are missing, so tell the model to quote any field that contains a comma and to check the column count on every row. And if product codes lose their leading zeros when you open the file, Microsoft says to open it through Excel's Import Text Wizard, which you can force by renaming the file from .csv to .txt, and set that column to text. Do not let the model touch image URLs it cannot see; Shopify says external image links must be publicly accessible.

If you are choosing which catalog or admin task to hand to AI first, the AI Strategy Sprint is a paid 3-day first step that starts with a free 15-minute AI brainstorm call.

How I'd approach this

Do it by hand with Claude the first two times and save the prompt that works. If you import the same supplier file every month, the next step is a small tool that takes the raw file, applies your rules, shows the Problems list and gives you a clean CSV, with no pasting. That is a short, cheap build, and it is worth doing once a supplier file is a monthly chore. See my guides to AI for small business and business process automation for how I think about repeat work. If a free habit is enough, I will tell you.