App Store aisles are full of “AI shopping assistants.” Most of them add a script, a monthly fee, and another thing that fights your theme for Core Web Vitals. Useful AI on Shopify is quieter: better search, better drafts for support, better internal tools for ops.
Storefront: search and merchandising first
Customers type the way they talk. Synonyms, typos, “that blue dress from the reel.” Retrieval over catalog + content beats a chat modal that cannot add a variant to cart. If you do add chat, ground it in live inventory and your return policy — and keep it off the critical rendering path. That sits with the rest of a disciplined Shopify build.
Back office: where the hours actually go
Drafting replies to “where is my order.” Tagging tickets. Suggesting a product description the merchandiser still edits. Summarising a week of reviews for the founder. Those jobs have a human in the loop by default, which is where LLM features belong until the evals are boring.
Do not outsource judgement to an app
Every extra app is latency, privacy, and a vendor who will train on your data if the checkbox is easy to miss. Custom apps and the APIs you already pay for are often the cleaner path. We treat AI on Shopify like any other integration: staging, review, and a kill switch.
Still deciding Shopify versus a custom stack? The commerce trade-off does not change because a model is involved.