How to Build Your Own Product Recommendation Engine With AI

Most product recommendations are chosen because two items share a collection or happen to be bestsellers.

That does not make them useful.

A strong recommendation anticipates what the customer needs next. It can increase order value, improve product discovery and make a large catalogue easier to shop.

Today:
We break down nine types of recommendation engines and show AI Launch Codes members how to choose and build the right ones using their Shopify data.

A customer adds one product to their cart.

You now have a brief opportunity to increase the order value. But only if the next product feels like a useful addition, not another sales pitch.

A product recommendation engine decides what each customer should see next based on the product they chose, their likely intent and where they are in the buying journey.

Done well, it can help you:

The problem is that most stores rely on generic recommendations:

“Customers also bought...”

“You may also like...”

“Shop our bestsellers...”

These suggestions fill the space, but they rarely answer the customer’s immediate need.

Someone buying sunscreen probably needs aftersun, not whichever serum happens to be your bestseller.

The strongest recommendation feels less like an upsell and more like helpful advice.

You would normally use an app such as Rebuy, Wiser, Glood or Selleasy to install these recommendations.

Plans can begin around $9 per month for a small store and rise into hundreds of dollars as order volume and features increase.

But before paying for another app, you can use AI to examine your catalogue and create the recommendation logic yourself in minutes.

You should not use all nine everywhere.

A fashion store may need visual matching. A supplement brand may need replenishment. A beauty brand may rely on functional pairings and sequential recommendations.

The right engine depends on what you sell, what you know about the customer and where the recommendation appears.

In today’s AI Launch Codes, we show you how each engine works, when to use it and how to generate its recommendations from your own catalogue.

This workflow was inspired by Louis Grommen’s breakdown of a smarter product recommendation system.

The complete workflow to build a custom recommendation engine for your store is at the end of this newsletter.

Join ADWEEK on Sept 16 at 1:00 PM ET for a live session digging into how top brands and agencies are approaching Black Friday and Cyber Monday. This session features research from Tatari, the CTV advertising platform that surveyed marketers across the industry on timing, budget, and channel strategy.

Highlights from the survey:


• 50%+ of brands are launching holiday campaigns by October
• 6 in 10 brands are increasing BFCM budgets this year

In addition to the survey results, this session will feature specific takeaways from a marketing leader at MANSCAPED, breaking down their channel mix and approach. If you want the playbook before the season hits, now's the time to grab a seat.

Nearly One in Three Teens Now Trusts AI Product Recommendations More Than Friends

Mastercard surveyed 26,000 parents and teens. For brands, visibility inside AI recommendations is becoming part of customer acquisition, especially for younger shoppers.

Amazon Confirms Prime Big Deal Days for October 6 and 7

Amazon will release new deals three times daily. Marketplace sellers should finalize inventory, pricing and promotions now, while DTC brands prepare for two days of aggressive discount competition.

Shopify Makes the Cart Permanently Accessible on Mobile POS

Store staff can now see cart contents and totals in one tap throughout checkout, removing a small but repeated source of friction for brands operating physical stores.

Treat the cart total as a promise. An unexpected £8.45 is not just an added cost. It changes the number after the customer has already decided to buy. Show the final total in the cart when possible. If delivery depends on location, show the likely range before checkout.

Open your store on your phone and add your most popular product to the cart.

Continue to the payment page without placing the order. Compare the price shown in the cart with the final amount the customer must pay.

Look for shipping fees, duties, taxes or other costs that appear only after the customer begins checkout.

Why this works:

A customer mentally accepts the price shown in the cart.

If that number increases during checkout, even a small charge can feel like the terms changed after they agreed to buy.

Your rep:

Report: “Final Price Checked”

One test checkout. One expensive surprise removed.