Shopify AI shopping in 2026: Catalog, agents, and the new battlefield

Shopify's own help page for this surface is published at help.shopify.com/en/manual/promoting-marketing/seo/optimizing-store-for-ai¹. The doc names AI-powered search engines and shopping assistants generically; the named channels — ChatGPT, Microsoft Copilot, Google AI Mode and Gemini, Perplexity, and Shop — appear inside the Catalog and Agentic Storefronts pages instead²⁵. That split matters: Shopify treats AI optimization as a content discipline and agentic distribution as a channel mechanic.

The same doc says, in Shopify's own words, that the merchant's job is to make product information "comprehensive" — detailed specifications, comparison information, comprehensive descriptions with relevant keywords, structured data and product attributes, sizing guides, material information, and care instructions, price, availability, and key features, plus high-quality images with descriptive alt text¹. None of those are AI-specific signals. They are the same things a good 2018 PDP had. What's new is that AI agents now read those fields directly and decide what to recommend on that basis.

Catalog eligibility has store-level and product-level requirements³. Store-level: the store must follow Shopify's Terms of Service and Acceptable Use Policy, be on the Starter plan or higher, not be password-protected, and ship to the United States or Canada. Products must be published to the online store, Hydrogen, or Headless channel; Agentic-plan users must include an external product URL. Product-level: a title, at least one product image, a price over $0, an identifiable product URL, no Unlisted status, no mature content.

Where the products go is documented across two pages. The Catalog page² lists "Shop, Select AI platforms, Shopping sites, AI agents, Agentic storefronts." The Agentic plan page⁷ names the AI channels explicitly: ChatGPT, Google AI Mode and Gemini, Copilot, Perplexity, the Shop app, with the trailing phrase "more coming soon." ChatGPT is described as a "discovery-focused referrer platform" — customers complete purchases on the merchant's storefront via an in-app browser or new tab — while the other channels support direct purchasing through Shopify-powered checkout inside the AI conversation⁵.

ChatGPT is the discovery-focused referrer⁵. OpenAI calls its consumer surface "Buy it in ChatGPT" and runs the Agentic Commerce Protocol with Stripe¹⁶. Copilot, Gemini, and Perplexity support direct checkout via Shopify-powered built-in checkout inside the AI chat. Shop is Shopify's own consumer channel and gets PDP block customisation features the others don't².

The two non-Catalog surfaces matter as much as the named five. Google AI Overviews lifts content from the classical Google index — a Shopify product has to be indexable, schema-marked, and answer-shaped to show up there. Claude reads Shopify storefronts the same way it reads any other web content; Anthropic publishes ClaudeBot for training, Claude-User for live retrieval, and Claude-SearchBot for the Claude.ai search surface. The Shopify-side answer for all of these is the same: every AI engine that recommends a product reads the product page itself before it reads Shopify Catalog's structured signals. Get the PDP right.

The official editable file is robots.txt.liquid, created in the theme's templates folder via the code editor and served at the store's root¹³. Shopify is explicit about the risk: "Incorrect use of the feature can result in loss of all traffic." Shopify Support will not help with edits to it. The dev docs reinforce that the Liquid version exists so that "default rules are updated regularly to ensure that SEO best practices are always applied"¹⁴ — a merchant who pasted a plain-text replacement gave up future updates.

The app's stated purpose is to "view and customize the FAQs that AI shopping agents use to answer questions about your store"⁹. It exposes Shopify-generated FAQs based on the store's settings and policies, monitors buyer inquiries, and lets merchants create custom FAQs⁹. FAQs are saved as metaobjects under Content > Metaobjects¹¹, which means they're addressable by Shopify-API consumers, not just by AI agents.

Auto-generated FAQs draw from four settings surfaces: Language settings, Customer account settings, Shipping and delivery settings, and Return rules¹¹. Auto-generated entries can be overridden but not deleted; "Return to default" reverts an override. Manually created FAQs are fully editable. The takeaway: cleaning up store policies first is how Knowledge Base earns its keep — the app is only as good as the underlying settings.

The plan includes adding products to Shopify Catalog, selling on agentic storefronts without an online store, and selling on the Shop channel⁸. What it doesn't include: a Shopify online store, Shopify Messaging, gift cards, manually-created orders, and Shop post-purchase offers. Agentic-plan stores must include an external product URL on each product, which is how the AI channel deep-links to the merchant's existing storefront when ChatGPT or Perplexity surfaces the product card.

The customer-visible flow is identical from the AI's side: a buyer sees a product card, taps it, and either completes checkout inside the AI chat (Gemini, Copilot, Perplexity, Shop) or lands on the merchant's existing storefront via an in-app browser (ChatGPT). The Agentic plan lets a non-Shopify brand show up in the same product cards as the full Shopify online stores it competes against. The full Agentic plan article covers what's lost in exchange.

Each field maps to a different Shopify admin surface⁴. Title and description are obvious; the rest aren't. Product organization (Type / Vendor / Collections / Tags) is the part most merchants leave to whatever the import tool guessed. Barcode is the GTIN field that links a Shopify product to the same product on Amazon, Google Shopping, and the offline retail world. Variants and Option name are how AI agents disambiguate "small" vs "medium". External product URL is the Agentic-plan deep-link target.

The implication: a Shopify store can be Catalog-eligible, robots.txt-correct, Knowledge Base-installed, and still not get recommended — because the title is "Hoodie" instead of "Heavyweight oversize cotton hoodie in stone", because the description is the manufacturer's stock paragraph, because Type/Vendor/Tags are blank, because the GTIN is missing. The product data hub walks the field-level rewrites.

Step one is mechanical — the Catalog requirements page³ answers it in five minutes. Step two opens the new admin surface at admin.shopify.com/agentic, where Shopify shows performance across all major AI channels, ranking queries, and product-data recommendations⁶. Step three opens the theme code editor and audits robots.txt.liquid. Step four runs through the seven product fields Shopify names. Step five — category capture — is the hardest one to fix and the easiest to misread: in some categories, two or three pharma or legacy brands hold 80-100% of AI citations, and a new entrant has to earn citations on third-party sites before AI surfaces consider it. The category capture article covers that case directly.

Instead of a 6-month agency retainer, it's seven days, fixed price, refundable. If you'd rather walk this yourself, every page on this guide gives you the same checklist a paid install would. The free Catalog audit reports back within 24 hours on the four diagnostic gates above and where the store currently stands.