Shopify Campaign Autopilot: What It Runs, What It Costs and Which Decisions to Keep

On 28 October 2026 Shopify made Campaign Autopilot available to all merchants, with more channels. A store sets a budget and some guardrails, and Autopilot reads the store, its products and its customers, builds campaigns across Meta Ads, Microsoft Advertising, Snapchat, the Shop app and Shopify Messaging, with X Ads to follow, decides which channels to run on and what to spend where, and moves budget towards what is working. It also suggests SEO changes for the store to approve.

For a brand that runs its own marketing, or pays somebody to, the question is what this hands over and what it costs. This is what Shopify’s own pages say Autopilot does, what it leaves alone, how it is billed, and, on our reading, which decisions a brand should keep however the campaigns are run. It belongs with the rest of growth and advertising across channels.

Shopify lists each channel and what Autopilot does on it. In Shopify Messaging it recommends email automations, such as abandoned checkout recovery and a welcome series, and one-time campaigns, which needs the Shopify Messaging app. Shop Campaigns find new customers on the Shop app and website, for stores that meet that programme’s eligibility. On Meta it runs Advantage+ campaigns on Facebook and Instagram, through a Meta Business Manager with billing and permissions completed. On Microsoft Advertising it runs Microsoft Performance Max campaigns, in regions where Microsoft Advertising operates, through the store’s own Microsoft Advertising account. On Snapchat it runs Dynamic Collection Ads from the product catalogue, and needs a Snapchat Business Manager, a live ad account in the store’s currency, a business address, a funding source and a Public Profile.

SEO updates are the one recommendation that needs no ad channel: page title and product description changes, reviewed with Sidekick and applied only once approved.

Shopify is explicit about the boundaries. Autopilot does not change existing manual campaigns, one-off emails built outside it, or channel set-ups the store manages itself, and it does not change product prices or store settings. On Meta in particular, Autopilot’s campaigns are separate from the store’s other Meta campaigns and use a separate product catalogue: they do not replace or merge with what the store runs through the Facebook & Instagram by Meta channel.

That separation is worth understanding before switching it on. A store already advertising on Meta would, on our reading, then have two sets of campaigns bidding for the same shoppers from two catalogues, and the numbers in each will need reading together, not side by side, to know what the extra spend added. Shopify’s pages also do not say how Autopilot decides when to move budget from one channel to another, so a store sees the result of each shift rather than the reasoning behind it.

Autopilot itself is free, Shopify says, with no fees; the store pays for the marketing it approves and runs. The bills come from different places. Meta, Microsoft Advertising and Snapchat each bill their ad spend directly, and none of it appears on the Shopify invoice. Email follows the store’s Shopify Messaging plan, and Shop Campaigns follow Shop Campaigns pricing and the campaign’s own settings.

The budget needs reading carefully. Shopify calls the monthly budget a spend target and says that in some cases Autopilot can spend more than it, because attribution across the connected paid channels is delayed. A brand with a hard ceiling should set its limits in each ad platform as well, where the platform enforces them, and watch the three platforms’ bills rather than one dashboard.

Autopilot is open on every Shopify plan except the Agentic plan. The store needs at least one product, shipping set up, the store out of private mode, good account standing, and staff with the Marketing permission. Shopify checks eligibility periodically, and if a store stops qualifying, Autopilot stops running its tactics.

To build campaigns it reads the active catalogue, with titles, descriptions, images, prices and stock; customer segments, purchase history and email subscriber lists; order history, revenue and conversion metrics; and the campaigns already running on connected channels. Data sent on to Meta, Microsoft or Snapchat is also governed by that platform’s terms once the channel is connected.

Autopilot lives under Growth in the Shopify admin. Setup connects the channels the store wants it to use, each through its own account and billing, and sets the guardrails it must work within; Sidekick can be asked to help with the setup and to review what Autopilot proposes. From then on the store receives recommended actions, email automations, ad campaigns or offers to win new customers, approves the ones it wants and rejects the rest, and can review performance and pause or stop any tactic or campaign at any time.

On our reading, connect one paid channel first, not all of them, so the first month shows what Autopilot does with one budget on one platform. Write down before switching it on what the store spent on that channel in the previous month and what it earned, so there is something honest to compare with, and decide in advance what result would make you add the next channel or switch it off.

An automated system works within the limits it is given, and the limits a store sets on the first day decide most of what it does afterwards. A short list written before setup saves corrections later.

Products to leave out. Products with thin margins, products often out of stock, and products that sell only alongside others. Advertising them costs money the margin cannot repay, or sends shoppers to pages they cannot buy from.

Offers allowed and not allowed. Whether discounts may be used, and how deep, and whether free shipping may be promised. A system optimising for sales will reach for a discount if it is allowed to, and a brand that never discounts should say so explicitly.

The words and the pictures. The tone the brand writes in, the claims it can support, and the images it is happy to see in an ad. Generated copy and creative should be reviewed against these until the brand trusts what comes out.

The money. A total monthly budget, and a ceiling per channel, set from what the store can afford to spend before it knows whether the spend works.

Who approves. One named person reviewing pending actions on a fixed day each week. Approvals left waiting are campaigns not running; approvals given without reading are decisions handed over without meaning to.

Autopilot can be set to need approval before it runs a tactic on a connected channel, in which case each one waits in Pending actions under Growth until somebody approves or dismisses it; once approved, Autopilot manages the campaign within the guardrails. On our reading, that setting should stay on, at least until the store has seen a few months of what Autopilot proposes, because the approval is where a brand’s judgement enters: which products are worth paying to sell, which offers protect the margin, and which audiences the brand wants to reach.

Three other decisions do not belong to any automated buyer. The first is what success means: a cost per order the margin can carry, not the platform’s own count of conversions, which is why incrementality testing matters more when the spend is spread across channels. The second is the channels themselves: Meta and Microsoft find different shoppers, a choice set out in Meta Ads against Google Ads for eCommerce, and Google is not on Autopilot’s list. The third is the calendar: peak season campaigns are built and approved weeks ahead, as in the Shopify BFCM checklist, not left to shift budget on the day.

Any system that runs a store's marketing also reports on how well it did, and each channel it uses reports its own conversions. Read alone, those reports describe the system's view of itself. A store should judge it on numbers the system does not control.

Total sales first. Compare the store's total sales, orders and new customers in the weeks after switching it on with the weeks before and the same weeks a year earlier. If the channels report many conversions and the store's totals have not moved, the conversions were mostly sales that would have happened anyway.

New customers against returning ones. Marketing that mostly reaches existing customers, who would have come back by email or by habit, looks efficient and grows the business slowly. The share of orders from new customers says whether the campaigns are finding new demand.

Spend across every bill. Ad spend arrives from several platforms, each billing directly. Add them together each month, set against the sales they are credited with, so the whole cost is visible in one place.

Test what it adds. Where volume allows, a period or a region without the automated campaigns, compared with one with them, shows what they actually add. Incrementality testing covers how to run that comparison.

Set the measures before switching it on, and read them after a full month, not after a good week.

Automating a store's emails and messages makes it easy to send more of them. The rules on who may be contacted, and how often customers will tolerate it, do not change because a system is sending.

Consent comes first. Marketing emails in the United States must offer a working way to unsubscribe and honour it; in the EU and the UK, marketing generally needs the customer's active consent in the first place. Marketing text messages in the US need the customer's prior written consent. An automated campaign sent to a list collected without the right consent carries the same risk as a manual one.

Keep the lists honest. Only customers who agreed to marketing should receive it, and an unsubscribe in one place should take effect everywhere the store sends from. Check that the automated flows read the same consent records as everything else.

Mind the frequency. Welcome series, abandoned checkout reminders, product recommendations and campaigns can stack up for one customer in one week. Too many messages raise unsubscribes and complaints, which harm delivery to everyone else.

Watch the signals. Unsubscribe rates, spam complaints and falling open rates are the early signs that the volume is too high. Read them monthly and reduce the sending before the inbox providers reduce it for the store.

On our reading, Autopilot fits a store that is not advertising yet, or runs only email, and wants campaigns on several channels without building each one, provided somebody reads the approvals and the bills. It fits less well a store that already runs Meta campaigns it understands, because Autopilot’s run beside them rather than inside them, or one whose margins need tighter control than a spend target gives.

We run paid media for brands across those channels and Google, with the account in the brand’s name; Meta Ads is one of them. If you are weighing Autopilot against what you run now, tell us what you spend and where, and we will tell you plainly whether it is worth switching on.