Shopify collection pages, rewritten at scale.
An established Shopify store had hundreds of thin collection pages. We built a script that wrote, reviewed, and published a unique page for every collection through Shopify's Admin API. Within months, those pages drew about twice the sessions.
Shopify PartnerAdmin APIClient name withheld
At a glance
At a glance · Seven months after launch vs. the same months a year before
- 01 / Sessions
- About 2x
- the sessions that started on a collection page
- 02 / Revenue
- 40%+
- more revenue from orders that started on a collection page
- 03 / Orders
- About ⅓
- more orders from shoppers who landed on a collection
- 04 / Scale
- Hundreds
- of collections, each with its own copy, meta title and meta description
05 / Shopify PartnerBuilt by a Shopify Partner, using Shopify's Admin API to publish at scale.
The challenge
The right products. Pages that weren't doing their job.
Thin collection pages
Many collections were just a product grid with a heading. Search engines had little reason to rank them, and shoppers arriving from search had little to confirm they were in the right place.
Too many pages to write by hand
Writing hundreds of unique pages by hand would have taken months, and the catalog would have changed by the time it was done. A shared template paragraph would have been fast, but it reads the same on every page and does nothing for search.
An aging theme with no version control
The theme was several versions behind, and years of edits had been made straight to the live theme. Every update was a risk.
An old domain still in search results
Listings from a previous domain were splitting attention from the current store.
How it works
One prompt per collection. Published through the Admin API.
Instead of one generic prompt, the script built a separate prompt for each collection from multiple data points, pulled together into optimized context. Then it wrote, reviewed and published.
- In.01Products in the collection
- In.02Product attributes
- In.03How shoppers find it
- In.0nMore data points
- Backup every collection
- Review output
- Publish via Admin API
# per-collection prompt · generic illustration collection: {{ collection.title }} products: {{ collection.products[0..n] }} attributes: {{ product.attributes }} shopper_intent: {{ how_shoppers_find_it }} context: {{ optimized_context }} write → body_html, meta_title, meta_description review → checks pass? yes → publish
What we did
Built a unique prompt for every collection
Generated copy is only as good as its inputs. The script built a separate prompt for each collection from multiple data points: its products, their attributes, and how shoppers find it. Two collections in one store read nothing alike.
Generated, reviewed and published at scale
The script wrote copy, a meta title and a meta description for each collection. We backed up every collection first, then the script reviewed the output and published it through Shopify's Admin API. It reruns as the catalog changes.
Where Shopify's built-in AI stops
Shopify's built-in AI is useful for a page here and there. But it works from what you give it in the moment, without the store-wide data that makes each collection different, and it isn't built to write, check and publish hundreds of collection pages in one pass. That's the kind of limit we run into on almost every project, and getting past it creatively is a big part of what clients hire us for.
The collection page
From a heading and a grid to a page worth ranking.
Abstract wireframe. No client content shown.
What else we did
A store the team can run on its own.
Redesigned the collection layout
We reworked the collection templates with category banners, a brand logo carousel and grid layouts, so shoppers see the right products right away. On mobile, the header tucks away so search stays in reach.
Upgraded the theme and moved it into Git
We brought the theme to its latest version and kept the store's customizations. It now lives in Git: we work on a development theme, the client's team reviews it, and it's published only after they approve.
Cleaned up what slowed the store down
- Set up 301 redirects from the old domain and resubmitted the sitemap
- Audited the installed apps for overlap and performance cost, and recommended which to replace or remove
- Compressed images, tightened code and improved load times
- Watched session recordings to see where shoppers got stuck, then simplified navigation, product pages and the path to checkout
Kept the in-house team in control
The store has its own team that runs the site day to day. Everything we built (a current theme, tracked changes, reusable templates and promo layouts) is something they can work with on their own.
The results
The results · Seven months after launch, year over year
We compared the seven months after launch with the same months the year before. Comparing year over year evens out seasonality and big product releases.
About2x the sessions that started on a collection page
More shoppers start on a collection page.
More shoppers now land straight on the right shelf instead of the home page.
- 40%+
- more revenue from orders that began on a collection page
- About ⅓
- more orders from shoppers who landed on a collection
More results
Organic search did the heavy lifting.
Out-earned its own ad
Early on, one rewritten collection page drew more visitors from organic search than the store's own paid ad brought in for the same search. That's traffic the store doesn't pay for, and it keeps coming after ad budgets change.
Those visits turn into revenue
Revenue from orders that began on a collection page grew by more than 40%, and orders grew by about a third. Better collection pages brought in more of the right shoppers, and those shoppers bought.
The gains were broad.
Traffic grew across many collections, not just a few. Some of the biggest gains came from collections that barely got traffic before, and a few new collections went from nothing to steady traffic.