Did your parents ever have a favourite child?  

No?  

Well, your ad algorithm definitely does.  

A merchant on Reddit watched it happen in real time:  

“I started a new e-commerce brand a week ago, mostly womens clothing (will be adding a lot of men's clothing soon) of approx 100 products women and 35 men clothing items... The PMAX (no assets, no text assets, feed only) is only showing (mostly) just jackets!!”  

If you’re wondering if this is a beginner mistake or a one-off glitch, it’s not. This is how most catalog ads work now.  

Over the past year, ad platforms have pushed harder into AI-based automation. Consequently product feeds now do more heavy lifting than most merchants realize. Which also means that the algorithm can scale what looks good at the account level while starving products that might deserve more budget.

So if your catalog ads have been working on autopilot and the results are worse than last year, in this article, we discuss the why. Let’s break down what actually changed, why algorithms obsess over the same handful of products, and how you can regain control and increase profitability.  

What actually changed about catalog ads this year

Imagine someone rearranges your entire store overnight and doesn’t tell you? One merchant experienced something similar when they logged in to find their catalog ads had become almost entirely automated

“I don't get who okayed this, it feels like this is the next step towards us handing over our URL or ad assets and them doing whatever they want with them…”

And just like that the control you wished to have running your own business is out of the picture.  

Your product feed now writes your ads

You know how Google Translate used to butcher everything? You’d type in a perfectly normal sentence and get back something that was technically the right words but completely wrong in spirit?

Something similar is happening to your catalog ads right now. Meta's Advantage+ Creative reads your product descriptions (the ones that a tired employee created using AI) and builds entire ad creatives from them. Things like backgrounds, moods, and sometimes even headlines. Google’s Performance Max does the same for Display and YouTube.

And despite all the recent conversations around “tastemaxxing,” AI has absolutely none of it.

A coffee brand describes their blend as “a dark, explosive roast” and the algorithm gives it a fiery, chaotic background that screams action movie, not artisanal morning ritual like you wanted.  

And you, a discerning merchant who puts thought into the smallest things about your brand, never approved this.  

Your audience targeting is basically a suggestion now

Meta rolled out its Andromeda update across ad accounts in late 2025. Its job is pretty simple - it reads your creative, analyzes your product feed, absorbs behavioral signals from billions of users and then decides, largely on its own, who should actually see your ads.
‍
One merchant on Reddit described its aftermath like this:

“I’ve been managing five websites since 2020. I’ve seen good times, bad times, technical headaches, scaling issues, but never have I lost as much money as I have since the end of this summer……Meta completely broke something. The update literally destroyed my performance overnight. CPMs went up, CTRs tanked, conversions vanished.”

They ended the post by saying:

“So don’t burn yourself out mentally trying to fix something that’s out of your control. Take a break, let ideas come to you, read a book, breathe a little”

Gosh.  

And to top it off, Meta has now started phasing out manual targeting controls from Advantage+ catalog ads altogether.

The gaps in your catalog ad setup that are leaking money  

Look at the following image:

Here's what's happening:

A man drops his keys in a dark parking lot but only searches under the streetlight because that’s where he can see. This is called “streetlight effect” and it’s how most merchants run catalog ads.

You might not know which products in your catalog are profitable

Meta will show you spend per product ID. It will not show you revenue or ROAS per product. Google’s Performance Max has the same gap. Most merchants and teams have started to identify this disconnect:

“Right now I allocate ad spend based on conversion rate and revenue. That is probably not the same as allocating it based on which products actually benefit from more spend. Some of that budget is probably working against me.”

The only way to get that number is to export your spend-by-product data from Meta, pull your sales-by-landing-page report from Shopify, and cross-reference them manually in a spreadsheet.

There usually isn’t a clean native report that tells you true product-level ROAS end to end. In Google’s Performance Max, you can see account or campaign-level value, but product-level spend-to-revenue matching still needs your own reconciliation.

How many merchants do you think are doing that every week?

Your campaigns run for months with a significant chunk of budget going to products that lose money on every sale. But nobody ever gets to notice it because the overall campaign ROAS looks fine.
‍
These ad platforms seemed to have introduced unwanted friction to businesses’ accessing their own key metrics. So you keep spending more on ads while optimizing around numbers that may have very little to do with real profitability.

The algorithm is playing it safe with your budget

Automated campaigns are fundamentally risk-averse. They find a few products with strong conversion signals and pour everything into them. One agency found that in a client’s Performance Max account with 2,278 products, 1,347 had never received a single click.

And let’s be real here, if a salesperson ignored half of your inventory, pushed a handful of products to every customer, never even attempting to sell the rest, would you call that incompetence or optimization?
‍
Marketers call these “zombie SKUs.” They exist in your feed, they’re live on your website. But inside your ad account, they might as well not exist.  

Every catalog ad looks the same

People have been saying it for quite some time now:

“Spent today going through skincare ads. A lot of them. Almost all saying the same thing. "Fresh ingredients." "Made with love." That whole world. Doesn't land. And the reason isn't complicated. Every single one was talking about the product. Nobody was talking to the person looking at the ad.”

Establishing your brand’s identity may have taken months of hard work to get a distinct appearance and story with hues that made your heart race. But when Meta’s AI takes your product image, overlays it on a generic lifestyle background, and adds auto-generated copy, none of that matters.  

Your ad is hyper-personalized but the landing page is not

If your prospective buyer clicks a catalog ad because they loved a very specific angle, maybe the styling, durability, or lifestyle context of a product, and then lands on a generic product page that reflects none of it, the journey breaks instantly.

Shopify’s own data shows that personalized CTAs convert 202% better than generic ones. The ad side of catalog advertising figured out personalization years ago. But your landing page never got the memo.

So, what’s the fix?

No, you don’t have to let the algorithm run your business into the ground. The fix actually lies in a handful of operational things you can start doing this week.

Audit true product profitability  

Stop trusting campaign-level ROAS. Export your Meta and Google ad spend by product ID, cross-reference it against Shopify product revenue or margin data, then reallocate budget based on product profitability, not just campaign ROAS.

Rescue zombie products instead of ignoring them  

Pull your zero-impression SKUs out of Performance Max and drop them into a separate Standard Shopping campaign with a Maximize Clicks bid strategy. By doing this you’re paying to force Google to actually test these products and collect initial conversion data. Once a product proves it can convert, move it back.

On Meta, the parallel move is creating dedicated product sets for untested SKUs so they’re not competing against your proven heroes for the budget. Give them their own ad set. Let the algorithm evaluate them on a level playing field.

One important caveat though. Not every zombie product deserves saving, some genuinely have low demand. The point is to separate “the algorithm never tested it” from “the algorithm tested it and it tanked.”

Write feed descriptions as briefs  

Treat your feed copy like an image generation prompt. Replace abstract marketing jargon with literal, sensory language (“rich dark roast in a warm kitchen setting”) that algorithms can translate without hallucinating the backgrounds.

There are tools that let you apply branded visual treatments across thousands of SKUs without designing each one manually. Overlays, frames, lifestyle photography at the product level. The goal is to make your catalog ads not look like catalog ads.

Bridge the landing page gap  

Never break the promise that starts with a hyper-personalized ad. Use tools like dynamic text replacement to automatically match your product page headlines to the exact UTM parameters and visual hooks that earned the click in the first place.

To conclude

The dream of the “set-it-and-forget-it” e-commerce business died the second platforms like Meta and Google rewrote their automation engines.

Right now, autonomy should be about resisting an algorithm to run your brand on autopilot while you bear the financial consequences. You must know exactly when to step in. Because in this new ecosystem, your product feed is the central nervous system of your entire business. It dictates how Meta’s Andromeda engine scores your creative, decides how Google displays your brand, and determines whether you exist at all inside the massive wave of AI-driven agentic storefronts.

Stop letting a risk-averse algorithm choose its favorite child based on lazy data. Take back your data, restructure your feed, and force the platforms to optimize beyond their own vanity metrics for your actual bottom-line profitability.  

‍