Shopify AI Product Recommendation Email With n8n and Gemini

You already know that repeat and add-on sales are where the margin lives, yet most stores stop talking to a customer the moment the order confirmation goes out. A Shopify AI product recommendation email built in n8n fixes that gap without a paid upsell app. Every new order triggers a workflow that reads what the shopper bought, gathers other products from the same category, and lets Google Gemini choose the three that pair best, then emails them a clean “you might also like” pick.

Prefer to skip the setup? Grab the ready-made template below and be running in under ten minutes.

A Shopify AI product recommendation email built in n8n turns every new order into a personalized cross-sell. When an order arrives, the workflow reads what the customer bought, pulls other products from the same category, and lets Google Gemini pick three complementary items to feature in a follow-up email. It runs on free tiers, sends through Gmail, and never invents products, because it can only choose from your real catalog.

What it does

The workflow watches your store for new orders. The instant one is placed, it looks up the product that was purchased, finds sibling products in the same product_type, and hands that shortlist to Gemini with one job: pick the three items that complement the purchase and write a one-line reason for each. Those three products get dropped into an email template and sent to the customer through Gmail.

The result is a recommendation email that feels hand-picked rather than random. Because the candidate list is scoped to a single category and comes straight from your Shopify catalog, every suggestion is a product you actually sell and can actually ship.

┌──────────────────────────────────────────────────────────────┐
│  SHOPIFY AI PRODUCT RECOMMENDATION EMAIL                      │
│                                                              │
│  [Order placed]                                              │
│       ↓                                                      │
│  [Shopify Trigger] → [Get purchased product] →               │
│       [Get same-category products] → [Build candidates] →    │
│            [Gemini picks 3] → [Build email] → [Gmail send]   │
└──────────────────────────────────────────────────────────────┘
  

Why it beats the default

Shopify’s built-in “related products” block on a product page is passive. It only helps a shopper who is already browsing, and it disappears the moment they check out. Paid recommendation apps close that gap but add a monthly fee and often a percentage of revenue on top.

This n8n approach reaches the customer in their inbox, right after a purchase, when intent is highest and the brand is fresh in mind. You control the category logic, the email copy, and the send timing. There is no per-order fee, no shared revenue, and no third-party script slowing your storefront. For the broader picture of what else you can wire up this way, see our guide to Shopify automation with n8n.

What you need

• A running n8n instance, either n8n Cloud or self-hosted.
• A Shopify store with Admin API access. If you have not connected n8n to Shopify yet, follow our 2026 Shopify to n8n connection guide, which covers the current Dev Dashboard method.
• A Google Gemini API key on the free tier, using the gemini-2.5-flash model.
• A Gmail account connected to n8n for sending the emails.

Build time is around 30 to 45 minutes from scratch, or under ten minutes if you import the template.

Node-by-node list

Seven nodes carry an order from placed to recommendation email.

1. Shopify Trigger fires on the orders/create topic the moment a new order is placed.
2. An HTTP Request node reads the purchased product so the workflow knows its product_type.
3. A second HTTP Request node pulls other active products in that same product_type.
4. A Code node trims the list into clean candidates, removes the item that was just bought, and attaches the customer name and email.
5. An HTTP Request node calls Gemini and asks it to pick the three best matches as JSON.
6. A Code node parses Gemini’s answer and builds the email HTML.
7. A Gmail node sends the recommendation email to the customer.

Step-by-step build

1 Shopify Trigger (orders/create)

Add a Shopify Trigger node and set the topic to Order Created. Attach your Shopify Access Token credential. This node hands the full order object to the rest of the workflow. The fields that matter here are the first line item’s product_id and the customer block.

{
  "id": 5123400012345,
  "customer": { "first_name": "Emily", "email": "emily.rodriguez@gmail.com" },
  "line_items": [
    { "product_id": 7781234500001, "title": "Cedar & Sage Candle", "price": "28.00" }
  ]
}

2 Get purchased product (HTTP Request)

Order line items do not include the product category, so fetch the product directly. Use an HTTP Request node with the GET method and your Shopify credential as a predefined credential type.

GET https://YOUR_STORE.myshopify.com/admin/api/2026-04/products/{{ $json.line_items[0].product_id }}.json

The response gives you product.product_type, for example Candles, which becomes the filter for the next step.

3 Get same-category products (HTTP Request)

Now pull a shortlist of sibling products. Another GET request filters the catalog by the category you just found and limits the pool so Gemini has a focused set to rank.

GET https://YOUR_STORE.myshopify.com/admin/api/2026-04/products.json?product_type={{ encodeURIComponent($json.product.product_type) }}&status=active&limit=15

4 Build candidates (Code)

Add a Code node to turn the raw product payload into a tidy list. It strips each product down to the fields Gemini needs, drops the item that was just purchased, and carries the customer details forward.

const purchasedId = $('Shopify Trigger').item.json.line_items[0].product_id;
const products = $json.products || [];

const candidates = products
  .filter(p => p.id !== purchasedId)
  .slice(0, 12)
  .map(p => ({
    title: p.title,
    handle: p.handle,
    price: p.variants?.[0]?.price || "",
    type: p.product_type
  }));

const c = $('Shopify Trigger').item.json.customer;

return [{ json: {
  firstName: c.first_name,
  email: c.email,
  purchased: $('Shopify Trigger').item.json.line_items[0].title,
  candidates
} }];

5 Gemini picks three (HTTP Request)

This is where the AI earns its keep. Add an HTTP Request node with the POST method pointing at the Gemini endpoint. Set the body type to JSON and paste a JSON body that includes the candidate list and a strict instruction. Attach your Gemini key as an HTTP Header Auth credential using the header x-goog-api-key.

POST https://generativelanguage.googleapis.com/v1beta/models/gemini-2.5-flash:generateContent

{
  "contents": [{
    "parts": [{
      "text": "The customer bought: {{ $json.purchased }}. From ONLY this JSON list of products, choose the 3 that best complement the purchase. Never invent products. Return JSON array of {title, handle, reason}. Products: {{ JSON.stringify($json.candidates) }}"
    }]
  }],
  "generationConfig": { "responseMimeType": "application/json", "temperature": 0.4 }
}

6 Build the email (Code)

Gemini returns its answer nested inside the API response. This Code node digs it out, parses it, and assembles simple product cards into an HTML email body.

const raw = $json.candidates[0].content.parts[0].text;
const picks = JSON.parse(raw);
const store = "https://your-store.com/products/";

const cards = picks.map(p =>
  `<p><a href="${store}${p.handle}">${p.title}</a> - ${p.reason}</p>`
).join("");

const first = $('Build candidates').item.json.firstName;

return [{ json: {
  to: $('Build candidates').item.json.email,
  subject: `${first}, three things that pair with your order`,
  html: `<p>Hi ${first},</p><p>Thanks for your order. You might also like:</p>${cards}`
} }];

7 Send with Gmail

Finish with a Gmail node set to Send. Map To to {{ $json.to }}, Subject to {{ $json.subject }}, and the message to {{ $json.html }} with the email type set to HTML. Save the workflow and switch it on.

Common mistakes

Cost at realistic volume

Take a store doing 600 orders a month. That is 600 recommendation emails, each making one Gemini call on gemini-2.5-flash. The Gemini free tier comfortably covers that volume, so the AI cost is zero. Self-hosted n8n is free and one order triggers a handful of lightweight executions. Gmail sends within its normal daily limits at this scale.

Compare that to a paid recommendation app charging a monthly fee plus a slice of attributed revenue, and the workflow pays for itself on the first recovered add-on sale.

Frequently asked questions

Related guides

• Shopify post-purchase follow-up email with n8n
• Shopify customer win-back email with n8n
• Shopify repeat customer reward email with n8n
• Browse every store recipe in the Shopify automation category.