Shopify's Native AI Merchandising vs Your App Stack: Keep, Kill, Replace
Summer Editions 2026 did something the app ecosystem has been dreading for years: it folded merchandising into the core Shopify admin. Three features matter. AI Collection Sort reorders collection pages automatically. Predictive Cross-Sell Blocks drop recommendation widgets into product and cart templates with no app embed. And a new Merchandising Insights panel surfaces the collection-level analytics you previously needed a third-party dashboard to see.
If you pay for a sorting app, a recommendation engine or a merchandising analytics tool — and most seven-figure brands pay for at least two of the three — some of that spend became redundant the day this shipped. Not all of it. The interesting work is figuring out which part, and that is what this post is for.
per month on merchandising-adjacent apps
Across the audits we run, a typical seven-figure DTC brand carries three to five apps touching sort order, recommendations or merchandising analytics. Most were installed by different people, at different times, for reasons nobody can fully reconstruct.
We covered the subscription side of this Editions release separately. This piece is about the merchandising tools, because they are the ones that put real monthly app spend in question.
What Shopify actually shipped, and where it stops
Before cancelling anything, be precise about what the native tools do. Shopify's launch copy and the actual behavior are not the same thing, and the gaps between them are exactly where your audit decisions live.
AI Collection Sort
Continuously reorders collection pages against an objective you choose, weighting recent sell-through, stock position and on-site engagement. No schedules, no rule trees to maintain.
The limit: it works at collection level, on storefront signals only. Pinning and exclusion controls are coarser than mature sorting apps, and it knows nothing about what a shopper did in your emails.
Predictive Cross-Sell Blocks
Theme blocks for product and cart templates that surface AI-selected complements. They inherit your theme's styling and ship zero third-party JavaScript.
The limit: recommendations live on the storefront and nowhere else. Segment logic is thin: you cannot build "returning VIP who bought X but not Y" rules, and you cannot reuse the model in email or ads.
Merchandising Insights
An admin panel showing which collections, grid positions and products actually carry revenue, including dead-weight products squatting on premium positions.
The limit: it is descriptive, not operational. It shows the problem but pushes fixes nowhere, and custom segmentation is minimal compared with a dedicated analytics tool.
The three merchandising features in Summer Editions 2026, and where each one stops.
Treat those limits as the audit's raw material. Every one of them maps to a legitimate reason to keep paying a third party. Everything not on that list maps to a reason to stop.
The keep/kill/replace audit
Start with an inventory, not opinions. For every app in your admin, capture four things:
-
Whether it touches sort order, recommendations, search results or merchandising analytics, including features bundled inside bigger apps you pay for anyway.
-
Its true monthly cost, including the usage tier you upgraded to during BFCM and never downgraded.
-
The last piece of evidence it produced measurable lift: an A/B result, a holdout test, anything with a date on it.
-
The native equivalent, and an honest answer to whether it covers 80% of the value.
Then run each app through three questions, in order.
Question 1
Can you point to measurable lift from this app in the last 90 days?
No
Kill
Cancel it regardless of what the native tool does. Unmeasured spend does not get the benefit of the doubt.
Yes
Continue
The app has earned the next question.
Question 2
Does the native equivalent cover at least 80% of that value?
Yes
Replace
Migrate to the native tool, then cancel. Follow the sequencing later in this post so the two never run at once.
No
Continue
There is genuine differentiated value. One question left.
Question 3
Is the remaining 20% worth the monthly fee plus the script weight?
Yes
Keep
Keep it, and diarise a re-test next Editions cycle. Keep is a verdict you earn with evidence, not a default.
No
Replace
Take the 80% for free and bank the difference. Most apps that reach this question still end here.
The keep/kill/replace decision flow. Run every merchandising-adjacent app through it in order — most don't survive question one.
"If you can't show the lift in ninety days of data, the app isn't underperforming — it's unmeasured. And in the audits we run, unmeasured spend defaults to kill."
The 80% threshold is deliberate. Native tools will rarely match a specialist app feature for feature, and they don't need to. You aren't paying for a feature list; you're paying for incremental lift over the free alternative. If native captures four-fifths of that lift at zero monthly cost and zero script weight, the app has to justify itself on the remaining fifth alone. Most can't.
Where third-party apps still win
Three categories consistently survive the audit, and it's worth being clear-eyed about why.
Cross-channel personalization
Native cross-sell blocks live on your storefront and nowhere else. If your recommendation engine also powers product picks in email flows, SMS and retargeting feeds from a single model, that is genuinely beyond what Shopify shipped. The lift in those channels is real and measurable, and cancelling the app forfeits it.
Deep segment logic
Quiz-driven recommendations, VIP-tier merchandising, B2B price-list-aware sorting, rules keyed to purchase history — the native tools have nothing like this. If your store's economics genuinely depend on showing different shoppers different products, a specialist engine still earns its fee. The honest follow-up question: do you actually run that logic, or did you buy the capability and configure two rules in 2023?
Search
AI Collection Sort does nothing for the search results page. Synonym handling, typo tolerance and merchandising rules applied to search results remain app territory, and for stores where a meaningful share of revenue starts at the search bar, that matters. In our experience, search-and-discovery apps survive the audit more often than any other category.
Notice what isn't on this list: on-store collection sorting and basic "you may also like" widgets. Those were the bread and butter of the category, and they are exactly what Shopify just absorbed.
The second saving nobody prices in
Every merchandising app you remove takes its scripts with it. Recommendation widgets are reliably among the heavier offenders in the theme audits we run: many ship their own rendering framework just to draw a product carousel, and sorting apps poll your storefront to keep their ordering fresh. That is main-thread work your shoppers pay for on every page view, and it lands directly on the metrics we broke down in our Core Web Vitals guide.
Native blocks render server-side inside your theme. No third-party JavaScript, no layout shift while a widget hydrates, no fourth copy of a carousel library. Replacing two merchandising apps with native equivalents typically removes several hundred kilobytes of script. We covered what app scripts actually cost your storefront in detail, and merchandising tools featured heavily.
One trap: uninstalling an app does not remove the code it injected into your theme files. Budget an hour to strip leftover snippets and script tags, or the performance saving never materialises.
What this looks like in money
Here is a composite stack from the seven-figure brands we audit: five apps, each plausible on its own, collectively carrying a lot of overlap.
App
Monthly
Verdict
Collection sorting app
Automated sort rules across 40 collections
$120
AI Collection Sort covers it
Recommendation engine
On-site widgets only; email module never enabled
$99
Cross-sell blocks cover on-store use
"Frequently bought together" widget
Overlapped the engine above for a year
$39
No measurable lift on record
Merchandising analytics dashboard
Collection performance reporting
$79
Insights panel covers the use case
Search & discovery app
Synonyms, typo tolerance, search merchandising
$249
22% of revenue starts at search
Before
$586/moAfter
$249/moSaved
$337/moA composite merchandising stack from seven-figure brand audits. Costs are illustrative; the overlap pattern is not.
Just over $4,000 a year. Not transformational on its own. But the canceled subscriptions are arguably the smaller half of the win. The lighter theme, the single source of truth for sort order, and the end of "which app is doing this?" debugging sessions are worth more than the line items.
The number that matters most isn't the saving. It's that two of the five apps failed question one entirely.
Migration sequencing: one system writes sort order
One failure mode reliably turns a tidy saving into a support ticket. Collection sorting apps work by switching your collections to manual sort and rewriting product positions through the API. Native AI Collection Sort wants to own the same field. Run both at once and they overwrite each other: sort order ping-pongs, your analytics become unreadable, and you'll conclude the native tool is broken when it's actually fighting your app.
The rule is simple: at any given moment, exactly one system writes sort order.
Baseline · week 1
Record conversion and revenue per session for your top twenty collections. Screenshot every app rule. Touch nothing. Without a baseline you cannot prove the migration helped or hurt.
Read-only native · week 2
Enable Merchandising Insights only. Compare its read on your top collections against the app's reporting; the discrepancies tell you which signals each system trusts.
Single-collection swap · weeks 3–4
Pick one mid-traffic collection. Pause the app's rules for that collection only, then enable AI Collection Sort on it. Two weeks minimum before judging the result.
Roll out in batches · weeks 5–6
Migrate remaining collections in batches, always pausing the app's rule before enabling native sort. Swap recommendation widgets for cross-sell blocks one template at a time.
Uninstall and strip · week 7
Cancel, uninstall, then remove leftover snippets and script tags from the theme. Re-run your speed tests against the week-1 baseline to bank the Core Web Vitals gain.
A seven-week migration sequence that guarantees only one system ever writes sort order.
The pattern, not the feature
Editions has a rhythm now: every cycle, Shopify absorbs another app category into the platform. It happened to bundles, to post-purchase offers, to chunks of checkout customization — and now to merchandising. Betting that your current app stack will still be the right shape in eighteen months is betting against a very consistent trend.
The brands that come out ahead aren't the ones reacting fastest to each announcement. They're the ones with a standing habit: every Editions release, re-run the keep/kill/replace audit against whatever the platform now does for free. The stack you assembled in 2023 was probably right for 2023. That's the strongest thing anyone can say for it.
And if you genuinely don't know what your merchandising apps are measurably contributing, that isn't a criticism — almost nobody does until someone goes looking. It's worth having someone look.