Bulk AI Suggestions

Two gaps stop an AI shopping agent from recommending a product more often than any others: a category that is too shallow, and category metafields that are empty. Bulk AI Suggestions closes both for up to 20 products at once. You select the products in My Store, press once and carry on working. The run happens in the background, you review every suggestion in one table, and nothing reaches Shopify until you press Save.

Figure 1. Nine products selected in My Store, filtered to Shallow Category. The green banner above the bar is an earlier run that is already waiting to be reviewed.

1. What It Covers

A run is a category pass. It looks at two things and only two things, because those are the two the agent uses to decide whether your product is even a candidate for a prompt.

Product category	The Shopify product category, taken as deep as the taxonomy allows. Outdoor Recreation becomes Ski & Snowboard Goggles.
Category metafields	Every metafield Shopify defines for that category, filled with a value Shopify accepts. Colour, material, width class, base material and so on.
Not in a run	Title, description, tags and image alt text. Those live in the Agentic Product Listing and the Product Listing editor, per product.

Note: the suggestions come from the same inference the single product editor uses, and they are written through the same code. A bulk run is the per product flow, run 20 at a time, with your review in the middle.

2. Start a Run

Open My Store and scroll to the product table.

Tick the products you want. The header checkbox selects everything on the page.
Press Bulk AI Suggestions in the bar above the table.
Carry on. You are not moved anywhere and you do not have to wait.

The button carries the whole lifecycle of the run, so it is worth reading.

Bulk AI Suggestions	Nothing running. Press it to start a pass on the products you have selected.
Suggesting…	The pass is running in the background. Selection is locked while it does, and the line to the left of the button says how many products are in flight.
View AI suggestions	The pass has finished and is waiting for you. Pressing it opens the review table.

Two limits apply, and the bar tells you when you reach either.

20 products per run	Selection cannot cross pages, so the page is the run. Select all 20 on a page, review, save, move to the next page.
5 open runs per store	Runs stack, so you can start a second pass while the first is still waiting for review. Once five are open, review or save one before starting another.

Tip: filter the table before you select. Filtering to the Shallow Category warning, then ticking the header checkbox, gives you 20 products that all have the problem a run is built to fix.

3. Follow It Without Waiting

A run reports itself in two places, both of which survive you leaving the page.

3.1 The Banner on My Store

One banner, one line per state, so the page never tells you two things at once.

Banner says	What is true	What to do
Bulk AI Suggestions: analyzing your products	The pass is still working. Products land as they finish.	Nothing. Come back later.
Bulk AI Suggestions: ready to review	Every product is done and nothing has been written to Shopify.	Press View products and review.
Bulk AI Suggestions: waiting for your review	A suggestion is parked against a product, waiting for a decision from you.	Approve it or discard it.

3.2 The Alerts Drawer

The bell in the app header holds one pinned entry for Bulk AI Suggestions, however many runs are open. It carries a real progress bar while something is moving and, when the work is done, a readout in units you can act on: how many products are approved and ready to be written, how many suggestions are parked, how many nobody has looked at yet. It ends with nothing saved yet, which stays true until you press Save.

4. Review Everything in One Table

The review surface is Bulk AI Suggestions, reachable from the banner, from the alerts drawer or from the bar's View AI suggestions. It is not an index of runs. It lists every product any open run is holding, in one table, so three products from three runs are three rows rather than three visits.

Product	Thumbnail, title and SKU. The thumbnail opens the product's listing page.
Current category	What the product's category is in Shopify today.
Suggested category	What the run proposes, as a picker. It wears the AgentIQ gradient outline only where the suggestion actually differs from the current category, so you can see at a glance which rows the model wants to move. Change it and the metafields below reload for the category you chose.
Metafields	How many of the category's fields would be filled, for example 12 of 13 filled. A field already set in Shopify counts as filled. Press it to expand the row.
Status	One badge, with the reason on the line underneath. See 4.2.

Above the table, one line counts the whole review: 0 approved · 1 pending · 8 not ready · 1 no recommendations, and under it the two things that line cannot say, Nothing is saved yet · Writes to Product Listing.

Figure 2. Every product a background run is holding, reviewable and saveable from one table, across every open run.

4.1 Field by Field

Expand a row, or press Expand all, to see one line per metafield of the suggested category. That includes the fields the model said nothing about, because 3 of 9 filled is only meaningful if you can see the other six. The expanded panel names the full category path it is working from, for example Category metafields for Apparel & Accessories > Clothing > Socks > Athletic Socks, and offers Approve all for the row.

Each field line carries four things.

A List badge	Every category metafield is a defined list, so the badge is the same on every line. It is telling you the field takes values from a fixed set, not free text.
The suggested value	As a chip with the AgentIQ treatment, and underneath it one line saying why, for example The description explicitly says the socks are made with merino wool. A field the model had nothing for shows an empty picker instead.
A value picker	Select a value, or Add another value where a value is already chosen. It lists only what Shopify accepts for that field.
Two buttons	A tick to accept the value, and a clock to leave it pending and decide later. Accept every field on a row and the row is approved.

Every value offered comes from your store's own Shopify definitions rather than being guessed. Colour values carry Shopify's own swatch, so Navy is Shopify's Navy and not the browser's. There is no free-text option, because a typed value has nothing in Shopify to resolve against and could never be saved.

Figure 3. One line per metafield of the suggested category, with the values Shopify accepts and nothing else. Three of nine filled here, and the six empty pickers are the point.

4.2 One Badge per Row, Nothing Mixed

A row is exactly one thing. Everything else worth knowing is a reason, and reasons sit on the line under the badge. Three badges decide what a save does.

Status	What Save does	Does the row leave?
Not ready	Nothing. The product is still being analyzed, could not be read, is missing data or has not been approved.	Stays
Pending	Stores the parked suggestions against the product. Nothing is written to Shopify.	Stays
Approved	Writes the category and the accepted metafields to Shopify.	Leaves

There is a fourth badge, and it is a result rather than a decision. No recommendations means the run looked and had nothing to propose, and the line under it reads Already up to date, approve to clear it. Approving such a row writes nothing. It is simply how you tell the table you have looked, and it is the only way to get the row off it.

Note: a product cannot be both. A row holding a parked suggestion, a value Shopify would reject or a field still loading cannot be approved, so one press can never write half a product and store the other half.

5. Save

Approving is per row. Saving is one press for the table, on Shopify's own save bar at the top of the screen.

Save N items	Writes every approved row to Shopify and stores every pending decision. Approved rows leave the table.
Discard	Drops the decisions you have made on this visit. The run itself is untouched.

If Shopify refuses a field, the save says so rather than reporting success: a warning on My Store names how many metafields were not accepted and how many products were updated anyway. The rest of the save still lands.

6. After a Save

A product you approved is written and gone from the table. A product whose suggestion you left pending stays in your catalog with the suggestion attached, reads Review pending in the My Store table, and comes back in the next run so the decision is never lost.

7. Good to Know

Needs review	The model proposed a value outside what Shopify allows for the field. The value is quarantined rather than written, the field renders empty and the row asks you to pick a value Shopify accepts. Nothing fails after you approved it.
Already set in Shopify	The field already holds a value. It counts as filled and the run leaves it alone.
Low confidence, worth a look	The model was unsure rather than wrong, and nothing is missing. The row can still be approved. The note exists so you can give it a second look first.
Could not be loaded	One product failed. The rest of the run is unaffected and still reviewable, and the failed row offers Retry.
Running it again	A second run on the same product replaces its earlier open row rather than adding a competing one.

Bulk AI Suggestions writes to your Product Listing, which is the record AI shopping agents read once your Agentic Product Listings are mapped into your Shopify Catalog.