Articles on: General

AI-Powered Product Recommendations with Autopilot

AI-Powered Product Recommendations with Autopilot


Choosing the right products to upsell or cross-sell can be time-consuming and challenging. That's where Autopilot comes in. This powerful AI feature automatically recommends relevant products for your offers, learning from your catalog and customer behavior. This guide walks you through everything you need to know about Autopilot and AI-powered recommendations.


Manual Selection vs Autopilot


When creating an offer, you have two ways to populate it with products:


Manual Selection

You personally choose specific products from your catalog. This gives you complete control but requires you to know which products work best together.


Autopilot (AI-Powered)

Let AI automatically suggest the most relevant products for your offer. Autopilot sources from your Collections, Tags, Product Types, or Metafields and uses intelligent algorithms to make smart recommendations.


When to Use Autopilot


Autopilot is ideal when you:


  • Have a large product catalog and manual selection becomes overwhelming
  • Want recommendations that adapt over time based on catalog changes
  • Need dynamic suggestions for each customer context
  • Want to reduce the time spent managing offer products
  • Have products organized by tags, collections, or metafields


Enabling Autopilot


Here's how to enable Autopilot when creating or editing an offer:


  1. Open the Offer Editor for your offer
  2. Click the Products tab
  3. You'll see the option to switch between Manual and Autopilot
  4. Select Autopilot
  5. Configure the AI settings (see next section)
  6. Save your changes


Once enabled, Autopilot takes over product selection for that offer.


The AI Config Tab


After enabling Autopilot, you'll configure it in the AI Config tab. Here's what each setting does:


Intent

This tells the AI what type of recommendations to make. Common intents include:


  • Related Products: Show similar items (e.g., other shirt colors/sizes)
  • Complementary Products: Show items that pair well (e.g., phone cases with phones)
  • Bundled Products: Show items that form a natural bundle (e.g., camera + lens)
  • Upgrade: Suggest premium alternatives to items in the cart
  • Cross-Sell: Show different category items (e.g., accessories with clothing)


Choose the intent that matches your business goal for that offer.


Max Products

Set the maximum number of products the AI will recommend. For example:


  • Set to 1-3 for focused, high-impact upsells
  • Set to 5-10 for more variety
  • Higher numbers work for cart drawers and product pages where space isn't limited


The AI will never recommend fewer products than needed to fill the offer, but will respect your maximum.


Inclusion Filters

Include only products that match specific criteria. You can filter by multiple attributes, and the AI will only suggest products meeting ALL filters. Options include:


Inclusion Filter Types


Collections

Limit recommendations to products in specific Shopify collections.


Example: For a "Winter Accessories" offer, include only products from your Winter collection. Or include products from both "Clearance" and "Outlet" collections.


Tags

Limit recommendations to products with specific Shopify tags.


Example: An electronics upsell might include only products tagged "electronics" or "tech". A gift bundle might include products tagged "gifts" or "bestseller".


Product Types

Limit recommendations to specific product types (as defined in your Shopify catalog).


Example: If you sell both apparel and footwear, an upsell on a shirt might only recommend products of type "Shirt" or "Accessory" but exclude "Shoes".


Metafields

Limit recommendations based on custom metafields you've created in Shopify. This is powerful for complex business rules.


Example: You might create a metafield called "product_line" with values like "budget", "premium", "enterprise". Filter to only recommend products with product_line = "premium" for a premium tier offer.


Combining Filters


You can use multiple inclusion filters together. All filters must be met for a product to be recommended.


Example: An offer might have:

  • Collections: Include from "Summer Collection"
  • Tags: Include only products tagged "womens" or "trending"
  • Exclude: Exclude products already in cart


This would recommend products that are in the Summer Collection AND (tagged womens OR trending) AND not already in the cart.


Using Metafields with Autopilot


Metafields are custom fields you create in Shopify to store additional product information. They're powerful for fine-tuning Autopilot recommendations.


Creating Product Metafields in Shopify


  1. In Shopify Admin, go to Settings > Custom data > Products
  2. Click Add definition
  3. Enter a name (e.g., "Recommended For", "Margin Tier", "Seasonal")
  4. Choose the data type (text, number, URL, etc.)
  5. Save
  6. Go to each product and fill in the metafield value


Linking Metafields to UpsellPlus


Once metafields are created and populated:


  1. In the AI Config tab, select Metafields as an inclusion filter
  2. Choose the metafield name
  3. Specify which values to include
  4. Save


Now Autopilot will recommend only products with matching metafield values.


Offer Types Supporting Autopilot


Autopilot works with most offer types:


  • Checkout Upsell: AI recommends products based on checkout context
  • Cart Drawer: Dynamic recommendations change as cart contents change
  • Product Page Upsell: Recommend related products when viewing a specific product
  • Post-Purchase Upsell: Suggest complementary items after purchase
  • Thank You Page Upsell: Show follow-up recommendations on order confirmation


Autopilot is less commonly used with "Checkout Header" offers since header space is typically limited, but it's still available.


Real-World Autopilot Examples


Example 1: Complementary Accessories


You run a tech store and want to suggest accessories when customers add electronics.


Autopilot Setup:

  • Intent: Complementary Products
  • Max Products: 3
  • Inclusion Filter (Collections): "Accessories"
  • Exclusion: Products already in cart


Result: When a customer adds a laptop, Autopilot recommends 3 accessories from your Accessories collection that pair well with laptops.


Example 2: Premium Upsell Based on Metafields


You want to upsell premium products only to high-value customers.


Autopilot Setup:

  • Intent: Upgrade
  • Max Products: 2
  • Inclusion Filter (Metafields): Include products where tier = "premium"
  • Smart Rules: Only show to customers tagged "vip"


Result: VIP customers see premium alternatives to items they're buying.


Example 3: Seasonal Cross-Sells


You want to recommend seasonal items based on what customers are buying.


Autopilot Setup:

  • Intent: Cross-Sell
  • Max Products: 4
  • Inclusion Filter (Tags): Include products tagged "summer" or "beach"
  • Inclusion Filter (Collections): Include from "Summer 2025" collection


Result: Customers buying summer items automatically see other summer products.


Example 4: Bundle Builder


You want to create dynamic bundles based on product properties.


Autopilot Setup:

  • Intent: Bundled Products
  • Max Products: 5
  • Inclusion Filter (Product Types): Include types "Shirt", "Pants", "Accessories"
  • Inclusion Filter (Metafields): Include products where bundle_group = "office_wear"


Result: Autopilot suggests 5 complementary items that naturally bundle together.


Monitoring Autopilot Performance


Track how Autopilot recommendations perform:


  1. Check Analytics for your Autopilot-enabled offer
  2. Monitor:
  • Conversion Rate: Are recommended products actually being added to cart?
  • Average Upsell Amount: What's the value of Autopilot-suggested items?
  • Impressions: Are the right products being shown to the right customers?
  1. Adjust intent, max products, and filters if performance is below expectations


Best Practices for Autopilot Success


1. Organize Your Catalog

Make sure products are properly tagged, placed in collections, and have metafields filled in. Autopilot is only as good as your catalog organization.


2. Start with Focused Intents

Begin with a single, clear intent per offer. "Related Products" or "Complementary Products" usually work best initially.


3. Use Meaningful Filters

Don't create filters that are too restrictive (only 1 product matches) or too broad (everything matches). Aim for 10-30 products matching your filters so Autopilot has options.


4. Test and Iterate

Create two similar offers, one with Manual selection and one with Autopilot, and compare performance.


5. Use Metafields for Complex Rules

If your business requires sophisticated recommendations beyond collections and tags, invest time in creating metafields.


6. Combine with Smart Rules

Use Smart Rules AND Autopilot together: Smart Rules determine when to show an offer, Autopilot determines what products to show.


7. Set Reasonable Max Products

Don't overwhelm customers. 2-4 products usually converts better than 10+ products.


A/B Testing Autopilot


Test Autopilot against Manual selections:


  1. Create Offer Version A with Manual product selection
  2. Create Offer Version B with Autopilot enabled
  3. UpsellPlus will automatically split traffic between versions
  4. Check Analytics after 1-2 weeks of data
  5. Scale the winning version


Troubleshooting Autopilot


Issue: Not enough products are being recommended

Check your inclusion filters aren't too restrictive. Make sure your catalog has products matching the filter criteria.


Issue: Wrong products are being recommended

Review your Intent setting. "Complementary" and "Related" recommendations are very different. Adjust filters to be more specific.


Issue: Customers aren't accepting Autopilot recommendations

Your products might not be well-matched to the context. Try a more specific Intent or adjust your Max Products number (try reducing it).


Conclusion


Autopilot transforms UpsellPlus from a manual product selection tool into an intelligent recommendation engine. By combining well-organized catalog data, thoughtful inclusion filters, and clear recommendation intents, you can create offers that genuinely delight customers while driving significant AOV improvements.

Updated on: 11/02/2026

Was this article helpful?

Share your feedback

Cancel

Thank you!