How to write Shopify product descriptions and metafields with AI, in bulk

Most catalogs have a long tail of products added in a hurry: one-line descriptions, empty metafields, SEO titles in five different styles. AI can fix all of that in an afternoon. It can also confidently invent a fabric, a dimension or a safety claim if you let it.

Here’s how to get the speed without the risk.

1. Decide what “good” looks like

Before touching any AI, write down:

• Your voice. A short paragraph describing how you sound, plus three product pages you love. Examples teach far better than adjectives.
• The structure. For example: one benefit-led opening sentence, a short paragraph, then a bulleted list of specs.
• The metafields that matter. The ones that power your filters and product pages, such as material, care instructions, dimensions or fit.

2. Export what you have

For a small catalog, Shopify’s product CSV export is fine. For larger catalogs, or when you need metafields too, use a bulk query in the Shopify Admin GraphQL API (bulkOperationRunQuery), which returns every product and its metafields as one downloadable file. Apps such as Matrixify are a good no-code alternative.

Pull in everything you know about each product: title, current description, product type, vendor, variant options, existing metafields and, if you have them, supplier spec sheets.

3. Give the AI facts, not freedom

This is the step that prevents trouble. The AI should only rewrite and organize information you give it. Tell it plainly:

Use only the facts provided. If a fact needed for a field is missing, write MISSING. Never guess materials, measurements, certifications or health claims.

A description that says “MISSING: care instructions” is far more useful than one that invents them.

4. Ask for structured output

Ask for the same fields for every product, in a fixed format, for example:

• description_html: the new description
• seo_title: around 60 characters, so it isn’t cut off in search results
• seo_description: around 155 characters
• metafields: one value per metafield you defined in step 1
• missing_facts: anything the AI couldn’t fill in from the data

Structured output makes the next step, review, quick and mechanical.

5. Review in a spreadsheet

Put the old and new versions side by side in a spreadsheet with an “approve” column. Start with a batch of 20 products, read them carefully, and adjust your instructions until you’re happy with the voice. Then run the rest.

Your team can now review hundreds of products quickly, approving the good ones and fixing the rest. Anything listed in missing_facts goes on a to-do list for whoever has the spec sheets.

6. Write it back, safely

Before you change anything, save a copy of the current values so every change can be rolled back.

• For a few hundred products, update them one at a time with the productUpdate mutation, at a pace that stays within Shopify’s API rate limits.
• For thousands, use a bulk mutation (bulkOperationRunMutation), which takes a file of updates and processes them on Shopify’s side.
• Metafields can be written with metafieldsSet, which handles several at once and creates or updates as needed.

Publish only the approved rows, and keep a log of what changed and when.

7. Translations

If you sell in more than one market, Shopify’s free Translate & Adapt app is the easiest place to start. For larger catalogs or a specific tone in each language, the same review-sheet approach works for translations too: generate, review, then write back through the translations API.

Common pitfalls

• Identical copy across similar products. Give the AI each product’s differences explicitly, or descriptions start to sound the same.
• Keyword stuffing. Write for people. Search engines reward pages that answer real questions.
• Regulated claims. Cosmetics, supplements, children’s products and anything electrical need a human check, every time.
• Forgetting new products. Keep the pipeline, so next season’s products get the same treatment on day one.

If you’d rather not run this yourself, a catalog cleanup is one of our fixed-scope sprints. Our free Shopify AI audit will tell you whether it’s worth it for your store.