Every Shopify store with a cart upsell faces the same quiet question: who decides what gets recommended? You can pair products by hand, one rule at a time. Or you can let software look at what is in each cart and suggest something that fits. The second approach is what people usually mean by AI cart recommendations.
This guide explains how they work, where they help, where manual rules still win, and what to check before you choose a tool. For a wider look at what good upsells look like, see our real cart upsell examples.
What Are AI Cart Recommendations?
AI cart recommendations are product suggestions shown inside the cart, chosen automatically based on what the shopper has already added. Instead of a merchant writing a rule such as "if this product is in the cart, show that one", the tool analyzes the cart and suggests complementary products in real time.
The goal is the same as any upsell: add something relevant to an order that was already going to happen, which raises average order value without paying for more traffic.
AI Mode vs Manual Rules
Most good tools let you choose between two modes, and the right answer is often a mix.
AI mode analyzes the cart and suggests products automatically. You do not need to know in advance which products pair well, and recommendations adjust as the cart changes.
Manual mode lets you define rules yourself, for example "if a passport holder is in the cart, recommend the travel case". Rules are checked in order, so you can stack several trigger and recommendation pairs.
Neither mode is better in every situation. They suit different stores.
When AI Recommendations Work Best
AI mode tends to fit when:
- You have a large or changing catalog. Writing rules for hundreds of products is slow, and the rules go out of date whenever you add or retire items.
- You are not sure what pairs well yet. Letting the tool suggest pairings is a faster way to find out than guessing.
- You want coverage without upkeep. Every cart gets a suggestion, even for products you never wrote a rule for.
When Manual Rules Still Win
Manual rules are the better choice when:
- You already know the best pairings. A skincare cleanser and its matching moisturizer, or a camera and its memory card, are pairings you do not need software to discover.
- You want to push specific products. If you need to move a high-margin item or clear stock, a rule puts it in front of the right shoppers.
- You have strict constraints. Some products must never be suggested together, and a rule enforces that exactly.
Many stores use AI mode as the default and add manual rules for the handful of pairings they care about most.
AI Recommendations in Different Places
The same idea shows up in more than one spot on a store, and the placement changes what is recommended.
- In the cart. Recommendations based on everything currently in the cart. This is where shoppers are closest to buying.
- On product pages. "Frequently bought together" suggestions based on the product being viewed, with an option to add the whole set at once. See our frequently bought together guide.
- In bundle building. Suggestions that help a shopper finish a bundle they have started, covered in how to set up bundles that convert.
Each can run on AI or on rules independently, so you can mix them to suit different parts of the store.
What to Look For in an AI Upsell Tool
Not every tool that uses the word "AI" gives you the same control. Check these before you commit.
- Relevance you can verify. Look at the suggestions it produces on your own products. If they look random in a preview, they will look random to shoppers.
- Manual override. You should be able to add rules and exclusions on top of the automatic suggestions.
- Control over how many suggestions appear. One to three well-matched items usually works better than a crowded row.
- A way to measure it. The tool should show how much revenue the recommendations earned, not just how many people saw them.
- Speed on mobile. A slow cart drawer loses the shopper before the suggestion appears. Test on a phone with a normal connection.
- Transparency about changes. You should always know what the tool is showing and be able to turn it off.
How Brix Handles It
In Brix, the cart upsell has an AI mode and a manual mode. In AI mode it looks at what is in the shopper's cart and suggests complementary products in real time, and you choose how many suggestions appear, from one to five. In manual mode you build your own trigger and recommendation rules. Frequently Bought Together on product pages and the suggestions inside the bundle builder work the same way, each switchable between AI and manual on its own.
The Brix AI assistant in the admin is separate. It helps you write upsell copy, set up a rule from a plain-language request, create a discount code and answer questions about recent store performance, and it always asks you to confirm before it changes anything.
Common Mistakes
Trusting automatic suggestions without checking them. Look at what shows up for your top ten products. If anything looks odd, add a rule or exclusion.
Showing too many options. A long row of suggestions asks the shopper to decide, and decisions slow people down. Keep it short.
Recommending items that are already in the cart or already bundled. Exclude products the shopper has chosen, and avoid repeating a bundle component as a cart upsell.
Never reviewing results. Check which suggestions get added and which get ignored every few weeks.
Changing everything at once. If you switch modes, change the layout and add a discount in the same week, you will not know what worked.
How to Test AI Recommendations
- Record your AOV for the last four to eight weeks as a baseline. Our guide on how to calculate and track AOV shows how.
- Turn on AI recommendations with a small number of suggestions.
- Wait two to three weeks of normal traffic.
- Compare AOV, conversion rate and revenue from recommended items to your baseline.
- Add manual rules for the pairings you know best, then compare again.
Conclusion
AI cart recommendations take over the job of deciding what to suggest, which saves effort on large or changing catalogs and finds pairings you might not have thought of. They do not replace your own judgment. The strongest setups let AI handle the broad coverage and use manual rules for the products that matter most to your margin or your brand.
If you do one thing this week, open your cart on a phone, add one of your best-selling products and see what gets suggested. If the suggestion makes sense, you have a good start. If it does not, you have just found the first rule to write.
Frequently Asked Questions
What are AI cart recommendations? They are product suggestions shown in the cart that are chosen automatically based on what the shopper has already added, instead of being set up one by one with manual rules.
Do AI cart recommendations increase average order value? They can, when the suggestions are relevant to what is in the cart, because they encourage shoppers to add a useful extra item. The size of the effect depends on your catalog and how well the suggestions fit, so measure it against a baseline.
Is AI better than manual upsell rules? Neither is better in every case. AI is useful for large or changing catalogs and when you are unsure which products pair well. Manual rules are better when you already know the best pairings or want to promote specific products. Many stores use both.
How many AI recommendations should I show in the cart? Usually one to three. Too many options slow the decision. Start small and increase only if the numbers support it.
Can I control what the AI recommends? It depends on the tool. Look for one that lets you add manual rules and exclusions on top of the automatic suggestions and lets you set how many suggestions appear.
Do AI recommendations slow down my cart? They can if the tool is poorly built. Test the cart drawer on a phone with a normal connection before and after turning recommendations on.