Shopify Smart Pricing, explained: how the AI price tips actually work

Pricing is the one number every merchant sets and almost nobody sets well. Too high and the product does not move; too low and you give away the margin that keeps the lights on. Most stores pick a price once, at launch, using a gut feeling and a competitor's screenshot, and then never touch it again. The money left on the table by that one-and-done habit is enormous, and invisible.

Shopify's Smart Pricing app, expanded as part of the Spring '26 Edition, is the platform's answer: an AI that watches your store's own data and suggests when to nudge a price up or down. This post is the neutral tour of how it actually works, so you (or the merchant you build for) can decide whether to use it. The harder question, whether you should trust each tip, gets its own developer's guide to reviewing AI price tips.

What Smart Pricing is

Smart Pricing is a first-party Shopify app that uses machine learning to recommend product prices. Per Shopify's Help Center, it learns each product's demand dynamics from your performance over time, your product and shop characteristics, and broader market trends, then suggests a better price when it sees one. The key framing from Shopify: "Any pricing changes are applied by you, so you remain in control of your product prices at all times." It is an advisor, not an autopilot.

It comes in two distinct modes, and this is the first thing to get straight, because a store only ever gets one of them.

The two modes

1. Price recommendations (markups and markdowns)

This is the everyday mode. Shopify looks at your store's inventory and sales, adjusts for seasonality and trends, and surfaces a tip: raise this price, lower that one. Tips refresh weekly. You review each one and apply it in a click, or edit it to whatever number you actually want, or ignore it. There is no experiment here: the tip is a model's best guess from history, applied immediately if you accept it.

2. A/B price experiments

This is the more powerful, more gated mode, in early access to select stores (US-focused at launch). Instead of trusting a model's guess, Shopify splits your live traffic between the current price and a test price and measures what actually happens to sales and conversion before you commit. It is real experimentation on real shoppers. Per third-party analysis from Intelligems, access has real requirements (a paid plan tier, a US market and single catalog, and a minimum number of products with enough monthly sales); treat the exact thresholds as a moving target and check the current docs.

What data it uses, and what it deliberately ignores

This is the most important section for understanding the tips you will see, because the model's blind spots are as defining as its inputs.

Shopify is explicit: "The model isn't trained on non-public price information, geographically localized information, or personally identifiable information (PII) about customers," and it does not share your confidential data with other merchants. That is reassuring for privacy. It also means the model is reasoning from your store in isolation: it cannot see that a competitor just undercut you, that demand is regional, or that one customer segment would pay more. Hold that thought, it is the whole subject of the companion post on when to overrule a tip.

The one nuance everyone misreads: store-level profit

Read this sentence from Shopify twice: "The Smart Pricing app optimizes for profit at the store level, rather than the product level." It means a tip on a single product is not necessarily the best price for that product. It is the move the model thinks is best for your store's total profit. That is why a batch of tips will contain both increases and decreases, not just increases. If you evaluate each tip in isolation ("why is it telling me to drop the price on my bestseller?") you will misjudge it. The model is playing the portfolio, not the position.

Which products get a tip

Not every product is eligible. Per Shopify, a product generally needs to:

• Have been created more than 30 days ago, with enough sales history to learn from.
• Not already be at what the model considers an optimal price.
• Have had no price or cost change in the last 30 days (a recent change resets the signal).
• Not have been in a pricing experiment in the last 3 months.

For products with multiple variants, Shopify generates a single tip that applies to all variants during an experiment, and you can exclude those products if per-variant control matters to you.

You stay in control

Three properties keep this from being scary. First, nothing changes automatically: you apply every price yourself. Second, tips are editable, a recent addition, so you can take the model's direction but set your own number (round to a .99, respect a MAP floor, whatever your strategy needs). Third, the adjustments are typically small nudges, not dramatic swings, and it does not reprice in real time per shopper. Shopify's own line: "you aren't obligated to change your prices in any way."

So, is it worth turning on?

For most stores that have never systematically revisited prices, yes, as a source of ideas. It is free, first-party, privacy-respecting, and it surfaces price moves you would never have spotted from your own dashboards. The recommendations mode is a low-risk way to get a weekly "here are five prices worth a second look." The A/B mode, where you can get it, is genuinely stronger because it measures reality instead of guessing.

But "turn it on" and "apply every tip" are different decisions. Every tip is a recommendation from a model that cannot see your competitors, your roadmap, or your brand strategy, optimizing a store-level number that may not match your goal for a specific product. Knowing when to accept, edit, or ignore a tip is the actual skill, and it is exactly the kind of judgment that stays valuable when the AI writes the suggestion. That is the next post.

Sources: Shopify Help Center, Overview of the Smart Pricing app; and Intelligems, Shopify Smart Pricing vs. Intelligems (third-party analysis of the A/B mode).