Shopify Sidekick Can Write One Meta Description. Here's What It Can't Do.

Ask Sidekick to write a meta description for one product and it does a decent job. It already knows the title, the body copy, the price, the collection the product sits in. It drafts something reasonable, shows it to you, waits for you to approve. Ninety seconds and you are done.

Now ask it which products in your catalog don't have one.

Sidekick is better than it gets credit for

It is worth being fair here, because a lot of the criticism aimed at Sidekick is lazy. It is not a chatbot bolted onto the admin sidebar.

Sidekick is included with every Shopify plan, with features and usage limits varying by tier. It reads your store data directly, writes ShopifyQL queries against it, drafts product copy and blog posts and emails, sets up discounts, and builds Flow automations from a plain-English description. It stages changes and asks for confirmation before anything is saved, which is the correct default and not every AI tool gets it right.

The Spring '26 Edition, shipped in June 2026, put Sidekick on every screen in the admin and opened it up to third-party apps through Sidekick app extensions. That last part matters for the rest of this post.

So the tool is real. The question is what shape of work it is built for.

The shape of the work

Sidekick answers inside a conversation. You point it at something, it looks, it responds, you approve. That loop is excellent for one product, one page, one campaign.

An SEO audit is not that loop. An audit is a crawl. Something has to walk every product, every collection, every page, and every blog article in the store, check each one against a set of rules, keep a record of what failed and why, and then hand you the result as a list you can work through. It is a background job with storage behind it, not a turn in a chat.

You can see the same assumption in how Shopify built app extensions. The developer docs describe data extensions as returning "records and counts" and cap tool responses at 4,000 tokens, with a one-second budget for data and three seconds for actions. Those are the dimensions of a conversational answer. They are not the dimensions of a catalog inventory.

None of this is a flaw. It is a scope decision, and it is a sane one. But it leaves a specific hole: the store-wide view.

Why the store-wide view is the one that changes anything

Writing one good meta description does approximately nothing for your traffic. The gains come from coverage, from the boring fact that 1,200 out of 1,400 products have a title and description that actually describe the product instead of falling back on whatever the theme generates.

Coverage is a counting problem before it is a writing problem. You cannot fix what you cannot list. And the merchants who most need this are exactly the ones who cannot eyeball it: if you have 90 products you already know which ones are thin, and if you have 4,000 you have no idea.

That is the gap a dedicated SEO tool fills, and it is the reason Seokai now ships tools into Sidekick rather than asking you to leave the conversation.

What Seokai adds to the conversation

Two read-only tools, both store-wide:

• An audit summary. The same health check Seokai runs across your whole store, returned as an answer in the chat instead of a dashboard you have to go open.
• A "what needs work" list. The entities that are missing metadata, so you get names rather than a percentage. Every result links back to the item in Shopify admin.

So you can ask Sidekick how your store's SEO is doing, or which collections are missing descriptions, and get an answer built from an actual pass over your catalog. Sidekick stays the interface. Seokai supplies the thing Sidekick cannot produce on its own.

The part that does not work, and why we are telling you

You cannot run an optimization through Seokai from Sidekick. The tools are read-only. Ask, get an answer, click through to the item, do the work in the app.

That was not an oversight. The optimize action was built and then deliberately parked. Sidekick overrides app deep links for Shopify-native resources, so asking it to "optimize my Contact page" makes Sidekick write its own SEO fields instead of opening Seokai. The handoff never lands. Shipping it would have meant a feature that appears to work and quietly does something else, which is worse than not having it.

If that behavior changes on Shopify's side, the action is already sitting there. Until then, read-only is the honest description.

Where this leaves you

The division of labor is fine. Sidekick is a capable hand on the item in front of you, and it will keep getting better at that. But "which 300 things are broken" is a different question from "fix this one thing," and it needs a tool that already went and looked.

Ask Sidekick to write the copy. Ask it what needs the copy first, and let something with a full pass over the catalog answer that one. If you want the fuller picture of how machine readers evaluate a store, agent readiness is the next thing worth reading.