Image Alt Text for Shopify SEO, Done at Scale

Alt text advice is usually written as though you have one image. Add descriptive alt text, include your keyword naturally, done.

A Shopify store with eight hundred products has several thousand images. At that scale, the advice is not wrong so much as unusable — a per-image workflow never finishes, and it decays continuously as new products arrive without it.

Treating this as a bulk problem is the only way it gets solved.

What good alt text says

The rule is simple and widely ignored: describe what the image shows.

Useful: "Navy wool overcoat on a model, showing length at mid-calf." Someone who cannot see it now knows what it depicts.

Useless: "navy coat mens coat winter coat buy coat online." This is keyword stuffing. It helps no reader, is transparent to search engines, and actively degrades the experience for screen reader users.

Also useless: "coat.jpg", or "product image", or an empty alt attribute left on a meaningful image.

The test worth applying: would someone who cannot see the image understand what it depicts? If yes, it is good alt text, and the relevant terms will appear naturally because you are describing a product that is what it is.

Different images need different descriptions

The failure that automation causes and that hand-writing avoids: giving every image of a product the same alt text.

A product with six images has six different pictures. The alt text should differ, because the images differ:

Six identical descriptions tell a screen reader user that they have looked at the same photograph six times, which is a worse experience than no alt text at all.

Decorative images should have empty alt

This distinction is routinely got wrong in the direction of over-describing.

Background textures, ornamental dividers, spacer images and purely decorative graphics should carry an empty alt attribute — alt="". That tells assistive technology to skip them entirely.

Describing decoration is not helpful thoroughness. It forces screen reader users to listen to "decorative gradient background" and similar, which is noise between the things they actually need.

The judgement is whether the image conveys information. If removing it would lose the reader something, describe it. If not, empty it.

Scale: rules for the tail, hands for the head

The workable pattern for a large catalogue is to split it.

Hand-written for commercially important products. Your best sellers, your highest-margin items, anything you are actively promoting. These deserve per-image descriptions, because a rule cannot know that image four shows the interior lining.

Rule-generated for the long tail. Templates built from product fields — type, colour, vendor, title — produce reasonable descriptions at volume. Not as good as hand-written, considerably better than nothing, and they apply to thousands of images in one operation.

Automatic for new products. This is the part that stops the problem returning. A rule applied on creation means new products launch with alt text instead of joining the backlog.

The preview step matters. A rule applied blindly across a catalogue can produce thousands of images described as "Untitled colour variant", and undoing that is more work than doing it properly the first time.

A worked example of building a rule template

Rule-based alt text works by combining product fields into a sentence pattern, and the difference between a usable rule and a poor one is usually in the small joins rather than the fields chosen.

A first attempt often looks like this pattern: product title, followed by colour, followed by product type. Applied to a product titled "Classic Overcoat" in the colour "Navy" and type "Coat", that produces "Classic Overcoat Navy Coat", which is readable but repeats information awkwardly when the title already contains a descriptive word close to the type, and it says nothing about which of the six images this particular one is.

A better pattern adds a connecting phrase and an image-position element: product title, then "in", then colour, then a phrase describing the image position such as "front view" or "detail" drawn from where the image sits in the product's image list. Applied to the same product, the first image produces "Classic Overcoat in Navy, front view" and the fourth might produce "Classic Overcoat in Navy, detail". This is not as specific as a hand-written description noting that the fourth image shows the interior lining, but it correctly signals that the images differ, which is the main failure the earlier section warns against.

The pitfall to check for before applying a rule across a whole catalogue: products whose title already includes the colour, which produces an awkward doubled phrase like "Navy Overcoat in Navy". A rule builder that lets you preview against a sample of products, including a few with unusual titles rather than only the tidy ones, catches this before it repeats across thousands of images. Products with missing or placeholder values in any field the rule references are the other common failure, since a blank colour field produces a sentence with a visible gap, such as "Classic Overcoat in , front view", which reads worse than having no rule at all.

The general principle carries beyond this specific example: a rule template is only as good as the data behind it, and the preview step exists specifically to surface the rows where that data is missing or unusual before the rule runs on all of them at once.

Why this matters more than it used to

Alt text has always been an accessibility requirement and a modest image search factor. Two things have increased its practical value.

Product data feeds more channels. Since Shopify's Summer '26 release, eligible product data syndicates automatically to several AI shopping surfaces. Those systems read fields. An image with no description is one less piece of information about what a product is.

Image-heavy catalogues carry information nowhere else. Many stores publish material, dimensions and detail primarily through photography. If the images are undescribed, that information exists only visually and is unavailable to every system reading the page.

Neither makes alt text a major ranking lever on its own. Both make it part of a data completeness picture that does matter, and that pays across classic search, shopping feeds and AI surfaces at once.

The accessibility case stands on its own

It is worth separating this from the SEO argument, because the SEO argument is modest and the accessibility one is not.

People using screen readers navigate a store through its text. On a product page, the images are the product — the colour, the fit, the texture, what it actually looks like. Without descriptions, a shopper using assistive technology reaches a page containing a title, a price and six unlabelled graphics, and has no way to evaluate what they are considering buying.

That is a customer who cannot buy from you, and in many jurisdictions it is also a compliance question. Accessibility requirements for commercial websites have tightened in several markets, and image descriptions are among the most basic expectations.

The practical consequence for how you approach this: write for the person, not the crawler. Every piece of advice above follows from that. Describe what is shown. Differentiate images that differ. Skip decoration. Do not stuff keywords. Each of those is an accessibility rule first, and each produces the better SEO outcome as a by-product.

Merchants who approach alt text as an SEO task produce keyword lists that fail both audiences. Merchants who approach it as description produce text that serves shoppers, screen readers, image search and AI systems simultaneously — which is a rare case of one piece of work satisfying every requirement at once.

The filenames question

Related and lower priority. Descriptive filenames — navy-wool-overcoat.jpg rather than IMG_4471.jpg — contribute modestly to image search and cost nothing when handled in bulk.

Worth doing as part of the same pass if your tooling supports it. Not worth a separate project, and not worth renaming files on a live store without checking what references them.

Two cautions specific to renaming. Images referenced directly in theme code, email templates or app configuration will break if their filenames change, and the breakage is silent — an image simply stops appearing somewhere nobody is looking. And renamed images lose whatever image search history the old filename accumulated, which for an established store is a real if small cost. Rename new uploads by default and leave established ones alone unless the current names are genuinely useless.

Image compression belongs in the same conversation. It rarely fixes a slow store on its own, since Shopify already serves images through its CDN with sensible defaults, but doing it during a bulk alt text pass costs almost nothing extra and removes one variable from any later performance investigation.

A sequence that works

1. Audit the gap. How many images have no alt text, and which products are affected.
2. Hand-write the products that matter. Usually a small fraction of the catalogue.
3. Build and preview rules for the remainder.
4. Apply in bulk, then spot-check the output.
5. Set rules to run on new products, so the gap does not reopen.
6. Handle decorative images with empty alt attributes.
7. Re-audit quarterly.

Step five is what makes this a solved problem rather than a recurring one. Without it, every alt text project is a one-off cleanup that begins decaying the moment it finishes, and you will be running the same audit again in a year.

For the wider product data picture, see SEO and AI search. Image weight and page speed are covered in storefront and design.