A merchant in menswear went looking for a filter. He had the products list open, the filter menu pulled down, and he wanted one thing out of it: show me the products with no photo. He wrote down what Shopify offered instead. Vendor. Tag. Status. Sales channel. Market. Product type. Collection. Publishing error. Gift card. Nothing about an image.
He asked the community on 26 April 2023 whether a no-image filter could be added. Nobody from Shopify replied, and he never posted on the thread again.
Two years and nine months later a second merchant found it and added to
it. He had solved the forward half, with a flow that tags any new
product missing_image when it arrives without one. What he
could not do was look backwards. Everything created before that flow
existed was still out there, and he had gone hunting on GitHub for a
script to export the whole catalog so he could find it in a spreadsheet.
That is the state of the art for a question that sounds like it should be a dropdown.
The short answer
The Shopify admin cannot filter on a field being absent. There are sixteen filters on the products list and none of them concerns an image, a weight or a barcode, and Shopify's documentation says the search bar and the filters do not support searching for products that are missing an attribute.
The product CSV will find a missing image and a missing barcode. It cannot find a missing weight, because a blank weight imports as zero and exports as zero, so nobody-weighed-this and this-weighs-nothing come out of the file looking identical.
Shopify Flow can guard the front door. It has a Product created trigger and a Product status updated trigger, and no Product updated trigger, so nothing fires when somebody opens a live product and deletes its last image.
The condition most people write first reports clean on exactly the products they built it to catch. A test written against each item in a list never runs when the list is empty, so the check needs the None operator rather than a question about a blank URL.
And an image that exists is not an image that will display. Shopify processes media asynchronously, so a product can hold media whose status is FAILED while any count-based check calls it fine.
What works is a gate plus a sweep: write down what finished means, hold anything short of it in draft, walk the back catalog once because no trigger covers it, re-check after the processing delay, and mail one short list every morning.
There is no filter for this, and Shopify says so
Shopify's page on searching for products and filtering your product list names sixteen filters: Vendors, Tag, Statuses, Categories, Sales channel, Region catalog, B2B catalog, Company location catalog, Retail catalog, Unassigned catalog, Types, Collection, Publishing error, Gift card, Combined listings and Metafields.
Read that list for its shape rather than its contents. Every filter on it selects products by a value they have. Not one selects products by a value they lack. The nearest miss is Publishing error, which tells you a sales channel has already rejected the product, which is the symptom arriving after the fact rather than the cause you could have caught at the keyboard.
The documentation is straight with you, which is more than most documentation manages. The same page says the search bar and filters do not support searching for products that are missing a specific attribute, and for the one case it addresses, products with no tags, its advice is to export the product list and read the column. Keep that sentence: it is the end of the first three hours most people spend here. There is no clever query syntax waiting to be found, and the export is not a workaround for a missing feature. It is the documented answer.
The CSV answers three questions and loses the fourth
So export. The CSV is a decent instrument right up to the point where it is not, and it helps to know in advance which point that is.
Images come out clean. A product with no image has an empty product image URL cell, and finding those is a sort. Barcodes come out clean too: Shopify's CSV reference describes that column as all of the barcodes for a product or variant, up to a maximum of twenty, and a blank cell is a blank cell. Worth knowing for when you refill them, a barcode supplied with no type is imported as a Custom barcode.
Weight does not come out clean, and this is the part nobody writes down. The same reference describes the weight column as the weight of the product or variant expressed in grams, and then states that the default value is zero if no value is entered. The file has no way to say unset. A pair of boots nobody weighed exports as zero. So does a gift card, and so does a digital download. They are the same row.
Which is why weight problems get found by customers rather than by audits. A zero is a legal value. It passes every import without a warning and quietly produces a shipping rate of whatever your cheapest band happens to be. Merchants meet it from the far end: a sticker and a winter coat costing the same to post, a free shipping threshold firing on orders it should not. The threads are always titled after the rates. The cause is a column that cannot tell absent from zero.
The Admin API can tell them apart, which is the useful half. An inventory item always carries a measurement object and the weight inside it is nullable, so null means nobody has said and zero means somebody has. The spreadsheet cannot carry that distinction. If weight matters in your store the CSV is the wrong tool, and you will keep reaching for it because the CSV is the tool the documentation pointed you at. It is the same gap that leaves variant metafields out of the file altogether. The format was built to move a catalog, not to audit one.
Flow will guard the door it has a trigger for
Shopify Flow is the right instinct and a partial answer. The product triggers are Product created, Product deleted, Product status updated, Product variant added, Product variant deleted, Product variant back in stock, Product variant out of stock, and Product variant inventory quantity changed.
Read that list the way you read the filter list. There is no Product updated.
So a product arrives with no photo and Flow catches it. A product is switched from draft to active with no photo and Flow catches that too. Somebody opens a live product, deletes the one image on it, saves, and nothing happens at all. The catalog is now wrong and no trigger fired.
Product created is narrower than it sounds, too. Shopify documents it as starting when a user in your organization or a third-party app adds a new product, in the admin or through the API, whether or not it goes to the Online Store sales channel, and then says it applies only to new products and does not start when a new variant is added to an existing product. So the new colorway on last season's jacket, added as a variant with no variant image, sits outside it. None of which makes Flow the wrong tool. It makes Flow the gate rather than the audit.
Why your first attempt misses the products you were looking for
This is the most useful paragraph here and it comes from no documentation at all. It comes from two merchants, three months apart, finding the same trapdoor.
In January 2024 a merchant with over two thousand items, many never photographed and some duplicated by staff without images, set out to build the check. The logic was simple in his head: if the product has no image, unpublish it. He wrote the condition the obvious way, as a test on the image URL, and asked which field to use, because the Product object offers featuredImage, featuredMedia, images and media and the builder gives no hint. He was told to use Product, then Images, then URL, and to leave Media alone. It worked. Three months later another merchant built the same flow and it did not run. His test product had no image at all, which was the entire point of the test, and the workflow stayed silent.
The reason will cost you an afternoon if you meet it alone. A condition written against a list runs once for each item in the list. If the list is empty the per-item test never runs, so a check for the image URL being empty finds nothing on a product with no images, because there is no image whose URL could be empty. What does run on an empty list is the Any, All or None operator. The working version uses None, phrased as none of the product's images existing, with the two action paths swapped so the empty case is the one that acts.
So the naive check does not merely under-report. It reports clean on precisely the products you built it to find, which is the worst failure a monitoring workflow can have, because it looks like good news. If you have one of these running today, test it against a product with zero images rather than a product with one bad image. Those are different tests and only one of them is the one you care about.
One more thing from that thread, because advice ages. The guidance that worked points at the images field and says to avoid media. On the current Admin API Product object, images and featuredImage are both marked deprecated and media is the live field, documented as the media associated with the product, where valid media are images, 3D models and videos. The Flow condition still works. It is not what you would write from the schema today, and anything you build against the API should go through media.
An image that exists is not an image that displays
Now the failure that survives a correct presence check, which is the one that catches careful people.
Shopify does not finish processing an image at the moment you attach it. Every media object carries a status and the enum has four values: UPLOADED, meaning the media has been uploaded but not yet processed; PROCESSING, meaning it is being processed; READY, meaning it is ready to be displayed; and FAILED, meaning media processing has failed. Media errors and media warnings sit alongside it.
Put that beside a check that counts images and the hole is obvious. A product whose single image has the status FAILED has an image. It has a row in the media connection, an id, a line in the CSV export. It shows a customer nothing. A count passes it and the storefront is where you find out.
Not a hypothetical. In August 2026 an agency-side thread went looking for how to verify product imagery after a bulk catalog change, on the stated grounds that image problems are harder to detect than text problems. The replies converge on one answer. Compare the intended state against the live state rather than before against after, recording ids, image references, positions, alt text and variant assignments and then querying them again. Treat a successful write as evidence the operation finished and nothing more. And, from the last reply, wait before verifying, because image processing is asynchronous: run the comparison at write time and again after a delay, and trust only the delayed result.
That is the discipline, written down by people who got it wrong first, and it is not new. In September 2021 a developer pushing tens of thousands of products into Shopify nightly asked for an API way to list the ones with no images so they could be fixed before customers noticed. He answered himself the same afternoon: check the response after creating each product, and if an expected image is missing, set the product to draft and send an email. The accepted reply a week later added that you should check whether the media had actually been processed. Five years on that is still the shape of the answer, and still something each merchant assembles alone.
What this costs on an ordinary Tuesday
Take a store with four thousand products and three people who add things. Finding the gaps by hand means exporting the catalog and sorting four columns. Call it forty minutes. You will do it once, feel good about it, and not do it again next month, which is how every catalog audit in recorded history has gone.
The cost of not doing it never arrives as a task. A product with no image converts at a rate you will never measure, because nobody emails to say they did not buy the thing they could not see. A product with no weight ships at the wrong postage until somebody in the warehouse notices the margin. A product with no barcode fails a channel feed, which turns up days later as a disapproval in an ads dashboard and gets handled by whoever runs ads rather than whoever created the product.
That is the real shape of this one. The damage is spread across people who did not cause it and arrives too late to connect back to the row that did. Which is exactly the kind of problem worth automating, and exactly the kind nobody puts on a sprint board.
What the automation actually has to do
As a Dugong playbook, in plain prose:
# trigger
Product created, variant added,
status changed, supplier feed
landed, plus a nightly sweep
# steps
1. Write down what finished
means in your store
2. Check media status, not
whether media exists
3. Treat a zero weight as
unknown, not as zero
4. Hold the publish instead of
reporting the damage
5. Sweep the back catalog once,
because no trigger covers it
6. Re-check after the delay,
because processing is asynchronous
7. Send one list every morning,
worst first
Step one is the one people skip and the one that decides whether the rest is useful. Finished is not a universal standard. A print shop needs an image, a weight and a description. A download shop would be wrong to require a weight. A store feeding a shopping channel needs a barcode on branded items and does not care on the ones it makes itself. Write the rule for your store, in a sentence, before anybody writes a query.
Steps two and three are the two checks everyone gets slightly wrong. Ask the media for its status rather than asking the product for a count, because one image with the status FAILED is a product with no image that happens to own a file. And treat a zero weight as a question rather than an answer: keep a short list of the product types that are genuinely weightless, usually one or two, and treat a zero anywhere else as unknown until a human says otherwise. That single rule turns the unanswerable CSV column into something you can act on.
Step four is where automation earns the difference from reporting. A report tells you about the damage, a gate prevents it, and holding anything unfinished in draft with a note about why is what that 2021 developer arrived at by himself. It beats a daily email because the broken product never reaches a customer.
Steps five and six are the two things Shopify will not do for you. There is no trigger for your existing catalog, so the back catalog is one large job that runs once, reports what it found and never needs to run again if the gate holds. And the sweep needs two passes, with the second one sending the mail, because the first pass is full of images that were merely still being processed.
Step seven decides whether anybody acts. A list of eight hundred incomplete products gets archived. A list of the nine added yesterday that are missing something gets fixed before lunch. Sort by what is costing money, which usually means active products with traffic, and send the short one.
The honest limits. The automation cannot weigh your boots or photograph your sofa, only refuse to let an unweighed, unphotographed product sell. For a catalog of forty products the spreadsheet is faster than describing a rule, and you should use the spreadsheet. And a gate will annoy somebody in week one, because sooner or later a product is wanted live now and finished later, so decide in advance who may override it.
Why this is a workflow and not a checklist
A checklist is a good answer when the work is rare. This work is not rare. Every product added, every variant added, every supplier feed that lands overnight, every bulk edit somebody runs at four in the afternoon is another chance for the catalog to drift away from finished. A rule enforced by a person who remembers is a rule that holds until the week gets busy.
And it has been asked long enough to stop waiting. The filter request is from April 2023 with no Shopify answer on it. The API question is from September 2021 and was answered by the person who asked it. The agency thread from August 2026 ends on good practice rather than a fix. Three forums, five years, the same hole.
With Dugong you describe this one in a sentence. A product is finished when it has at least one image whose processing succeeded, a weight somebody has actually entered, a barcode if the channel needs one, and a description; hold anything unfinished in draft and tell me what is holding it; sweep the whole catalog on the first of the month. That is the specification. Everything after it is plumbing, and plumbing is what automation is for.
Do this before you close the laptop
Three checks, ten minutes, no code.
Export your products to CSV and sort by the product image URL column. The number at the top of that sort is the number you did not know, and the surprise is the useful part.
Sort the same file by weight and count the zeros. Everything in that count is either a deliberately weightless product or a product nobody has weighed, and the file cannot tell you which. If the number is bigger than your list of genuinely weightless products, you have found your shipping rate problem.
Then, if you already have a flow checking for images, open it and point it at a product with no images at all. If it stays quiet, it has been sending you good news about a catalog it never looked at.
If you sell on Shopify and have a version of this story, the product
that sold for a month with no photo, the sticker that shipped at coat
postage, the feed that rejected four hundred rows on a Friday, the inbox
is open: field-notes@dugong.live. We are collecting case
studies for the next issue.