Ranked #1 for an AI query and invisible for another one
Tested real shopper queries against three Shopify catalogs
When a shopper asks ChatGPT, Copilot or Gemini to find a product, none of them browse your website in the first place. And for Shopify brands specifically, they query one index, called Shopify Catalog. It returns whatever ranks highest for that.
Here's the part that matters: Shopify Catalog runs primarily on structured attributes, the fields actually filled in on a product, not just text sitting in a description.
Shopify sends this data to AI agents, and of everything in there, only three fields are today available for an agent to filter on: color, size, and gender, and this isn't fixed - for more categories, more fields are expected to become filterable too.
Rank determines which results the agent actually shows the shopper.
ChatGPT and other agents each decide how many products to surface in a response, and it's usually a small handful, not a scrollable list - so if you're not near the top of that ranking, there's a real chance the agent never shows you at all.
I ran three brands through Atomz's Shopify Catalog audit and tested shopper queries against each one, to see how this plays out in practice.
If you haven't seen how the audit works, here's the full breakdown.
Ranking runs per product and per query, not per brand - we saw this with True Classic a few weeks back, where one product swung from #1 down to #5, outside the four results actually shown, depending on the query alone. Read it here →
This time it's three completely different catalogs and the range wider than I expected.
Lulu and Georgia
Home furniture and decor, 391 products.
Their descriptions run short across most of the catalog, and only about a third of it has color, size, and gender filled in as structured data.
The gap shows up once you start testing real queries:
→ "Hand knotted wool rug" - #1 to #4 (all of top four)
→ "Large sculpted wall shelf" - #2
→ "Cordless table lamp" - #7
→ "Table lamp" - #16, well past where an agent would typically pull from
→ "Pet friendly rug" - not in the top 50 at all
Query tested: Hand knotted wool rug
The closer a query sits to the product's name, the more likely those words are already in the description as text and the attributes - Shopify can infer and match.
Once the query turns generic, the kind of phrase a shopper types when they don't know exactly what they want yet - there's less for the system to grab onto.
Alo Yoga
447 products, and a very different starting point.
Almost every product has color, size, and gender filled in as structured data. Descriptions run long enough. By Shopify's own standards, that's a well-built catalog.
→ "White softsculpt tank" - #1
→ "Cropped yoga hoodie" - #4
→ "Tank top for workout" - #11, flagged
→ "High waisted leggings" - #26, flagged
Alo Yoga, "high waisted leggings"
"High waisted leggings" is the category Alo is arguably best known for, and it still lands at #26, behind Gymshark's $15 pair at #1 and Fashion Nova's $8 pair at #3.
The hoodie result shows why, in a way you can actually see on the page.
Alo's own listing has material down as "not specified", but the three products ranked above it all show a real material breakdown. So on that one query, Shopify simply has more to go on with the competitors than with Alo.
Fashion Nova
449 products, and a more mixed setup than either of the two above.
Color and size are filled in across the catalog but most descriptions run short, and target gender is barely structured at all. It's a different kind of gap than Lulu and Georgia's - not missing the basics, just thin on detail and gender.
→ "Grey jumpsuit" - three of the top four spots
→ "Long sleeve jogger jumpsuit" - #1
→ "Black comfortable jumpsuit" - #2
The first three are close variants of one sampled jumpsuit, so a win there isn't surprising. Two broader terms had unexpected results.
"High waisted jeans" is broad and generic. Fashion Nova still ranks #1.
Query tested: Bodycon dress
"Bodycon dress" is where real competition shows up - Oh Polly holds three of the top four spots, and Fashion Nova lands at #8. So a thinner catalog didn't cost them the win on one specific query, but it didn't protect them on the next query either.
What actually moved
All 3 brands won the query closest to the product's exact name, and that part tracks with how much data each catalog had to draw from.
Lulu and Georgia and Alo Yoga lost ground once the query turned generic. Fashion Nova broke that pattern - it won a broad, competitive term outright, then lost on another, even with thin descriptions and a real gap in gender structuring.
So a complete catalog helps, but it isn't a guarantee, and a thin one doesn't automatically lose either. What decides it, query by query, is whether that specific word or attribute is sitting somewhere in the catalog as real, AI-readable data.
This is exactly what Atomz fixes - fills in the structured attributes Shopify would otherwise have to guess at and expands thin descriptions straight from your existing catalog, so there's real data to match against instead of an inference.
- Ankit
If this was useful, the next one will be too.
Weekly. Free.
Unsubscribe with one click.
