Your products are shoppable inside ChatGPT. Your answers should be too.

Since March, millions of Shopify merchants have been able to sell directly to ChatGPT users through Agentic Storefronts. Your products can surface inside ChatGPT, Microsoft Copilot, AI Mode in Google Search, and the Gemini app — and for existing Shopify merchants, access arrives automatically through the admin. The catalog part is handled for you.

The answers are the part still worth your attention. When a shopper asks an assistant whether your jacket is warm enough for a Chicago winter, something has to supply that answer. Here's how to make sure it's you.

What Shopify turned on

Agentic Storefronts became publicly available in March 2026. It runs on the Universal Commerce Protocol — the open standard Shopify co-developed with Google for how AI agents transact with merchants, now supported by Walmart, Target, Etsy, American Express, Mastercard, Stripe, and Visa. Shop Pay is coming to Copilot, and UCP will power checkout in some Meta experiences.

Then in June, the Spring '26 Edition removed the approval requirement entirely. Any developer can register an agent profile and call the public MCP endpoint; the Catalog API dropped its access barriers and added image and multi-modal search. The layer went from invitation-only to open infrastructure in a quarter.

For a merchant, the practical summary is short: your products are now discoverable and purchasable in places you don't control, through interfaces you didn't build, in conversations you never see.

Agents answer from structured data — or they improvise

An AI assistant recommending products isn't browsing your site the way a shopper does. It's querying structured data. Prices, variants, availability, policies, and — where they exist — published questions and answers.

Where that data is rich and specific, the assistant answers precisely. Where it's thin, the assistant does what language models do: it produces the most plausible-sounding answer available from whatever it has. Product titles, a truncated description, category conventions, and general knowledge about similar products.

That second outcome is worth sitting with. A shopper asks whether your supplement is third-party tested. Whether the sofa fits through a 30-inch doorway. Whether the shoes work for someone with a wide foot. If the answer isn't published anywhere an agent can read, the shopper still gets an answer — assembled from context rather than from you.

Your product page has the same problem, and you already know it. Most brands have answered these questions hundreds of times in support tickets and DMs. The difference now is that the conversation increasingly happens before the shopper reaches your site, in an interface where you have no support inbox and no chance to correct the record.

Where AI-visible answers come from

Shopify gives merchants a native place to publish product Q&A as structured data. It's the same layer the Storefront MCP exposes to AI agents, and it's the answer to "how do I get my own words in front of these assistants."

That's the layer PlainSaid writes to. When you publish a product's Q&A, PlainSaid puts the answer in two places at once: the widget on your product page, where shoppers ask in the moment of decision, and Shopify's native Q&A metaobjects, where the agent ecosystem reads. Same answer, authored once, in your voice, grounded in facts you approved.

Three details matter for how this behaves in practice:

Only complete answers syndicate. PlainSaid surfaces questions shoppers have asked that you haven't answered yet — those are held back, never published. An agent should never represent your store with a blank, and a half-written answer is worse than no answer at all.

Entries are keyed to the product, not the handle. Product handles change when merchandising changes. The published answers stay attached through renames, so a URL edit in March doesn't quietly orphan your Q&A in June.

Edits propagate. Change an answer and it updates on the page and in the AI layer together. There's no second copy to maintain and no stale answer surviving somewhere you've stopped looking.

The questions worth answering first

Publishing everything at once isn't the goal. The questions that pay off in agent conversations are the ones a shopper can't resolve by looking at photographs:

• Fit and sizing specifics. Not "runs true to size" but the actual measured comparison — which of your other products it matches, what to do between sizes, how it fits a wide or narrow foot.
• Compatibility and dimensions. Will it fit the doorway, the bike, the older model, the standard mount.
• Materials and sourcing. Where it's made, what it's made of, what certifications it carries and which it doesn't.
• Care and durability. What happens in the wash, in the rain, after two years.
• Who it's for and who it isn't. The most underrated answer a brand can publish. Agents recommend confidently; telling them plainly who a product doesn't suit prevents recommendations that end in returns.
• Policy specifics tied to the product. Return windows, warranty terms, and shipping realities that differ by item.

A useful test: take the last twenty pre-purchase questions your support team answered. How many are published anywhere a machine can read? That gap is the work.

What to do this week

Check what an assistant currently says about your best-seller. Ask ChatGPT or Gemini three real pre-purchase questions about it. The answers tell you exactly where your published data runs out — and the exercise takes four minutes.

Publish answers to what shoppers already ask. If you run PlainSaid, the questions your shoppers have asked are already grouped by theme in the admin, gaps flagged. If you don't, start with your support inbox.

Write for a reader who can't see the page. An agent has no photographs and no size chart graphic. Answers that reference "the image above" or "see the chart" are invisible. Say the measurement.

Keep it current. Formulations change, suppliers change, policies change. An outdated published answer is more damaging than an absent one, because it will be repeated with confidence.

The shift underneath

For a decade, the discipline was making product pages that convince people. That work still matters — most purchases still involve someone looking at your page. But a growing share of the deciding now happens somewhere else, in a conversation between a shopper and an assistant, and the only thing you can send into that conversation is well-structured, accurate, specific information about what you sell.

Shopify has built the pipe and opened it to everyone. What travels through it is the part that's yours: your product knowledge, written down properly, in a form a machine can quote without inventing anything.

That's the same work that makes a product page good. It just reaches further now.