Shopify alt text

Shopify Alt Text

Backfill missing image alt text across a Shopify store. Alt text is an accessibility requirement first (screen readers announce it) and an image-SEO signal second. This skill fills the gap for images that have none; it does not rewrite alt text that already exists.

The sister backfill for SEO titles and descriptions is shopify-seo-metadata: same safe-mode doctrine, different fields. The FIND stage (measuring how many images lack alt across the catalog before you commit to a sweep) belongs to shopify-catalog-audit.

One mutation, two find paths

Images live in two places on a Shopify store, but both write through the same mutation. A product image and a Files-library image are both MediaImage nodes, which are File subtypes, so a single fileUpdate sets alt on either. There is no separate product-media alt mutation to learn.

• fileUpdate is the one write. It takes the image's MediaImage GID and the new alt. This is the write path both source tools converge on. (Shopify's old productUpdateMedia is deprecated on shopify.dev; do not use it for alt text.)
• Two ways to FIND the images, one way to fix them:
  • Product media: page products, read each MediaImage under product.media. Needs read_products.
  • Files library: page the files connection (Content → Files: theme sections, metaobjects, rich-text, collection banners). Needs read_files.
  • Both hand you MediaImage GIDs; feed those straight into fileUpdate.

Full count/fetch/update/verify recipes for both find paths are in references/queries.md.

Store access

Every skill that touches the Admin API opens with this stanza. Two lanes; pick one.

Lane A, custom-app token (scriptable). In Shopify admin: Settings → Apps and sales channels → Develop apps → create an app → grant the minimum scopes below, then install and copy the Admin API access token. Export it; never write it to a committed file:

export SHOPIFY_STORE="your-store.myshopify.com"
export SHOPIFY_ACCESS_TOKEN="<your Admin API access token>"   # keep it in the env, not on disk

Call the GraphQL endpoint (minimum scopes for this skill: read_products to enumerate product-image owners, read_files to page the Files library, and write_files because fileUpdate writes alt for both):

curl -s "https://$SHOPIFY_STORE/admin/api/2025-07/graphql.json" \
  -H "X-Shopify-Access-Token: $SHOPIFY_ACCESS_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"query":"{ shop { name } }"}'

Lane B, Shopify CLI OAuth (no stored token). shopify store auth --store $SHOPIFY_STORE --scopes read_products,read_files,write_files then shopify store execute to run a validated operation. Good for token-less stores where the owner logs in interactively.

Toolkit preflight. Lane B rides on the Shopify CLI: run shopify version first and install it if missing. For the full Admin GraphQL schema and validated execution, pair this skill with Shopify's official AI toolkit plugin. In Claude Code, check claude plugin list; if it isn't there:

claude plugin marketplace add Shopify/Shopify-AI-Toolkit
claude plugin install shopify-plugin@shopify-ai-toolkit

Recommended, not required: Lane A needs only curl and a token. The toolkit gives your agent the API; this skill gives it the playbook.

Recipes, not scripts

This is a recipe, not a program. The GraphQL blocks in the reference are copy-and-adapt starting points: write throwaway code per engagement against the current schema, run it, discard it. Do not ship a standing binary that silently rots when Shopify changes the schema.

Safe-mode / preview / verify doctrine

State it standalone, run it every time. This is a bulk write against a live store; the difference between a clean sweep and a support ticket is discipline.

1. Fill missing only, never overwrite. Treat an image as a candidate only when its alt is null or empty after trimming whitespace. An image that already has alt text, even mediocre alt text, is out of scope. Overwriting existing alt is a separate, deliberate decision a human makes per-image, never the default of a sweep.
2. Preview a read-only count first. Before any mutation, run the count query for the find path you are about to touch and read the number back. Sanity-check it: a store with 300 products and "1,900 product images missing alt" is plausible; "40,000 missing" on the same store means your filter is wrong (probably counting non-image files or already-captioned images). A count far above expectation means stop and re-scope, not proceed.
3. Preflight the generation API key with a real 1-token completion. If you caption images with an LLM, before the bulk run send one real (cheap, 1-token) completion and confirm a 200. Do not trust a models.list / auth-only check: a key that lists models can still fail on the actual completion (billing, per-model access, org limits), and you find out 400 images into the sweep instead of before it.
4. Log every write, then verify against the Admin API. Keep a per-write log (image GID, find path, old alt, new alt, status) so a bad batch is auditable and reversible. After writing, re-query the same images through admin/api/.../graphql.json and confirm alt is populated. The storefront is CDN-cached and shows stale (empty) alt for minutes; it is not evidence the write landed. The reference has the readback query.
5. Never print or commit the Admin API access token or the generation key. Read both from the env.

What good alt text is

Alt text describes the image for someone who cannot see it. It is not a place to stuff keywords or repeat the product's sales pitch.

• Describe the image, not the marketing. "Brown leather weekender bag with brass zipper, side view", not "premium handcrafted bag, free shipping, shop now."
• Keep it under ~125 characters. Screen readers read the whole string; long alt is punishing to listen to. Most screen-reader guidance treats ~125 chars as the practical ceiling. Cap generated captions at 125 before writing.
• No keyword stuffing. Repeating the product title plus a list of search terms hurts the accessibility use it exists for and reads as spam to search engines. One natural description.
• Skip decorative and duplicate images. A pure background texture or a spacer needs empty alt (alt=""), not a caption. If five images on a product are near-identical angles, they do not each need a distinct paragraph; differentiate only what a sighted user would notice ("front", "back", "worn on model").
• No "image of" / "photo of" prefixes. Screen readers already announce that it is an image; the prefix is redundant noise. Strip it if a generator adds it.

If you generate captions from the image (vision model), still gate every write through the safe-mode rules above. Canonical guidance: the W3C WAI alt-text decision tree.

References

• references/queries.md: count / fetch / update / verify recipes for both find paths (product media and the Files library), all writing through fileUpdate, with pagination and MIME filtering.

Provenance and maintenance

Last verified: 2026-07-05. GraphQL pinned to Admin API 2025-07; Shopify deprecates versions on a rolling quarterly schedule, so confirm fileUpdate against shopify.dev before trusting a version-specific claim. productUpdateMedia is deprecated; fileUpdate is the current alt-text write for both product media and Files. Read-only re-verification a stranger can run, listing product media and seeing which lack alt while mutating nothing:

curl -s "https://$SHOPIFY_STORE/admin/api/2025-07/graphql.json" \
  -H "X-Shopify-Access-Token: $SHOPIFY_ACCESS_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"query":"{ products(first: 5) { edges { node { title media(first: 10) { edges { node { ... on MediaImage { id alt } } } } } } } }"}'

Canonical docs: fileUpdate and the WAI images tutorial. These skills capture operational lessons the docs don't; they are not a replacement for the reference.