Bulk edit SEO titles, meta descriptions and alt text in Shopify

Three situations bring merchants to this page. The first is a store that grew faster than its listings: products were added from a supplier feed or a dropshipping app, the admin filled in the title and price, and the search engine listing fields stayed empty because nobody scrolled down. Google fills the gap by picking a sentence from the description, and the sentence it picks is rarely the one you would have chosen.

The second is a redesign or a migration. The theme changed, the collections were rebuilt, the product types were renamed, and the SEO titles still say what the old site said. The listings are not empty; they are wrong, which is worse, because the empty ones at least fall back to the product title.

The third is an audit. Someone ran a crawler, or Search Console flagged duplicate titles, or the Merchant Center review found thin descriptions. The report lists a few hundred products. Each one needs a sentence written, and the sentence has to be about that product. This is the job where "bulk" and "per product" have to be true at the same time.

In the Shopify admin, an SEO title and meta description are edited per product. Open the product, scroll to "Search engine listing", click "Edit", fill two fields, save. Alt text is one level deeper: open the product media, click the image, edit the alt text. There is no column for any of these in the admin bulk editor. For twenty products this is a coffee break. For four hundred it is a week nobody has.

A CSV export and import handles the volume. Export the products, fill "SEO Title" and "SEO Description" in the sheet, import. The sheet does not write the text; you do, or you build a formula that concatenates the product title and the vendor and hope the result reads like a sentence. The import has no preview and the admin has no undo. If the column shifts by one, the import runs anyway.

Table-style bulk editor apps put the SEO fields in a grid with a filter. That fixes the scale problem and adds a preview. It still leaves you typing the value, or applying a template like "{title} | {vendor}" to every row. Templates produce identical structures across the catalog, which is what Google's duplicate-title report is about. The missing piece in every one of these is a value written for each product from that product's own data.

Start with a question, not a change. Reads need no approval, and the answer gives you the number you will compare the preview against later. If the audit says 312 products have no meta description and the plan later says it will update 1,046, the scope moved and you decline.

The answer comes back as counts and a table: how many active products have no meta description, how many have no SEO title, how many images have no alt text, and the first twenty of each with product type and vendor so you can see what kind of products they are. Draft and archived products are excluded unless you ask for them; a listing nobody can see does not need a search snippet.

The instruction names the field, the scope and the rules. The rules matter more than the wording. Google shows about 60 characters of a title and about 155 of a description before truncating; Shopify stores up to 70 and 320. Say the limit. Say what must be in the text and what must not.

Scope is read literally. "Every active product that has none" leaves existing values alone and skips drafts. "Every product" would include the ones written by hand months ago. The difference between the two is the difference between filling gaps and overwriting work, and the preview will show you which one you asked for, but it is cheaper to ask for the right one.

ApiMate reads every product in scope first and states the count. Then it writes one value per row from the product title, type, vendor, tags and the existing description, and lists the rows. The list is the plan. It is not a sample: every row that will be written is in it, and a row that is not in it will not be written.

Two rows from a run, to show the shape: "Harbor Glove 1235 in Waxed Canvas | Kestrel Supply" and "Vantage Tent 1310 in Recycled Polyester | Basalt Works". Product name, material, vendor. Nothing that is not in the product data. A product whose description is empty and whose tags say nothing gets a shorter, plainer line, because there is nothing to say and the generator does not make things up.

The meta description is the two lines under the blue link in a search result. Google rewrites it when it thinks it can do better, which is often when the stored one is empty or generic. A specific one, naming the product, its material and its brand, is the one Google keeps.

The generator works from what the product has. Title and vendor are always there. Product type is usually there. Tags and the description body are the source for the specific detail: the material, the size range, the use. When the description body is a supplier paragraph about shipping, the generated line will be plain, and that is the right outcome; the fix is the description, not the snippet.

The rule "no exclamation marks, no free shipping" is not taste. Merchant Center reads the same fields and flags promotional text in them. A meta description that says "Buy now!!!" is a snippet Google may ignore and a feed row Google may reject.

The SEO title is the blue link itself and the browser tab. When it is empty Shopify falls back to the product title, which is fine for a product called "Merino Wool Beanie, Charcoal" and bad for one called "MWB-CH-01". The job is to fill the empty ones with something a person would click, under 60 characters, without producing four hundred titles that all end in " | Your Store".

The instruction for titles needs one extra rule: what goes at the end. "Product name, then the brand" is one pattern. "Product name, then the material" is another. Say it, or the generator picks, and it is easier to check a preview against a rule you wrote than against one you did not.

Titles that exist but run over 60 characters are a second scope: "rewrite the SEO titles longer than 60 characters, keep the product name first". The current column shows the long one, the new column the short one, and you decide per row whether the cut kept the right words.

The same flow works for image alt text, with one difference: the write goes to the image, not the product. ApiMate reads the product and its media first, then builds one row per image. A product with four images is four rows. The count in the scope line is the number of images, and that is the number to compare with the audit.

Alt text describes what is in the picture for someone who cannot see it, and for the image search index. "Charcoal merino wool beanie on a white background" is alt text. "Buy the best beanie" is not. Keep it under 125 characters, because screen readers cut off around there, and do not repeat the product title on every image of the same product; say what is different about this picture.

The generator does not look at the picture. It writes from the product title, type, options and the image position: first image is the product, later ones are likely details or alternate colours. If your images are unusual, the rows for a product will be similar to each other. That is a known limit, and the preview is where you catch it.

The preview is the whole job. Read it like a proofreader, not like a dashboard. Five minutes here is the difference between a catalog that reads well and a catalog you revert on Monday.

One approval writes every row. Shopify processes the writes in batches; a few hundred rows finish in under a minute, a few thousand take longer, and the job reports progress and the final count. When it finishes, the job appears in history with the rows it wrote and the values they replaced.

To see the result in Shopify, open any product from the list and scroll to "Search engine listing". The new values are there, and the preview panel Shopify draws under them shows the snippet as Google would render it. For alt text, open the product media and hover the image. Google will pick up the new snippets on its next crawl, which is days for a busy store and weeks for a quiet one; Search Console shows when.

Every write stores the value that was there before. Reverting a job builds a new job from those saved values and writes them back, through the same preview and approval, to exactly the rows the first job touched. For fields that were empty, revert makes them empty again.

What revert cannot do: restore a value that was changed by something else after the job ran. If you approved the job on Monday, edited three products by hand on Tuesday, and revert on Wednesday, the three hand edits are overwritten with Monday's saved values. Revert puts back what this job overwrote, not what the product looked like at any other time. The preview for the revert job shows every row, so the three are visible before you approve it.

Handles are not touched. A handle is the product URL; changing it breaks every inbound link until a redirect is in place. That is a different job with a different risk, and it is not bundled into an SEO field job.

Titles are not rewritten wholesale. The product title is the page H1, the name in the cart and the line on the order confirmation. Change titles one rule at a time ("append the material") with a preview, or not at all. An instruction that reads as "rewrite all titles" comes back as a question, not a plan.

Text comes from the product data. If the description is empty and the tags say nothing, the generated field is short and generic. The tool does not look up the product on the internet to fill the gap, and it does not copy text from another product. A thin catalog produces thin snippets; the fix is the descriptions, and that is its own case.

Alt text is written without seeing the image. For a catalog where the image order is consistent, the result is good. For one where it is not, expect to exclude rows.

There is no syntax. These phrasings all resolve to a plan of the same shape; the scope line at the top of the preview tells you whether they resolved to the scope you meant.