Merchant Stack Daily
Shopify app picks and stack advice for lean merchants

Shopify Sidekick Explained: What the AI Assistant Actually Does Well

What Sidekick is, mechanically

Sidekick is Shopify built in AI assistant, reachable from the admin search bar and the Sidekick panel, and it works against your store own data and Shopify documentation. Ask it for last quarter revenue by channel and it builds a report using ShopifyQL style queries under the hood. Ask it to draft a product description, a discount, or an email subject line and it generates from the catalog context you point it at. It can also take guided actions in the admin, walking settings changes step by step rather than guessing through menus you half remember.

That data grounding is the real distinction from a general chatbot. A general model knows what a typical store looks like. Sidekick can see your store: your products, your orders, your traffic sources. The reports it builds are only as honest as the question, though, and the question is where merchants quietly go wrong.

Ask badly, get a confident wrong number

Blended conversion rate is the classic trap. Ask Sidekick how conversion is doing and you will get a blended figure that hides the traffic mix problem underneath: paid social converting at a fraction of email while the blend looks stable. The operators who get value ask segmented questions by default: by source, by device, by new versus returning, with the comparison period named. Sidekick is fast enough that asking the segmented version costs nothing. The blended version costs decisions.

Same discipline for actions. Let it draft the discount, then read the discount. AI assistants stacking a percentage on top of an existing automatic discount is not a hypothetical failure mode, it is a Tuesday. Draft fast, verify like an auditor, publish slow: that order of operations is the entire management system for this tool.

Where it genuinely saves hours

Report building for one off questions that used to die in a spreadsheet queue. First drafts of product descriptions in a consistent voice when you feed it two good examples. Theme content edits described in plain language. Digging a setting out of the admin you have not touched in a year. Customer segment sketches you then refine in the real segmentation tool. Notice the pattern: retrieval, drafting, and navigation. Sidekick is a speed tool for work you already know how to judge.

Where we do not hand it the keys: pricing decisions, anything touching checkout logic, and customer facing sends that skip a human read. The cost asymmetry is the argument. A slow report costs minutes. A wrong price or a mangled discount costs margin at machine speed.

The habit that separates value from noise

Keep a running list of the questions you ask weekly and turn the good ones into saved reports and real dashboards, because an answer you need every Monday should not depend on a chat thread. Use Sidekick to explore, use the boring saved report to operate.

And verify one number per answer against a source you trust, every time, until you have calibrated how it handles your store data. Trust in an assistant, human or otherwise, is a measurement, not a mood.

Related reading

If this topic is useful, these earlier guides go deeper on connected questions: Shopify Metaobjects Explained: Custom Data Without a Custom App, Checkout Extensibility Explained: The Blocks That Replaced checkout.liquid, and The Best Free Shopify Apps Worth Installing.

Frequently asked questions

Is Shopify Sidekick free?

Sidekick is built into the Shopify admin for eligible stores and plans as part of Shopify Magic tooling, rather than sold as a separate subscription. Availability of specific features varies by plan and region and has rolled out in stages, so check the Sidekick panel in your own admin.

Can Sidekick change my store settings?

It can guide and initiate admin tasks, such as drafting discounts, building reports, and navigating settings, with your confirmation. Treat every proposed action like a draft from a junior employee: read it in full before approving, especially anything involving prices, discounts, or customer facing content.

How accurate are Sidekick reports?

They are computed from your store data, so the arithmetic is real, but the framing is yours. A blended metric can be accurate and still mislead. Ask for segmented views, name the comparison period, and spot check one figure against a saved report before acting on it.

What should I not use Sidekick for?

Pricing strategy, checkout customization decisions, and unreviewed customer facing sends. Use it to draft and to explore in those areas, and keep a human decision and a verification step between its output and anything a customer or your margin can feel.