When AI Agents Do the Shopping, Proof Has to Travel With the Product

For twenty years, online trust has been built on the storefront. A clean site, a returns policy in the footer, reviews under the product photo, a page explaining how to spot a fake. All of it assumes a person is looking.

That assumption is starting to crack. On 8 September 2026, Meta launched Muse, a personal agent that can find products for people. The same day Shopify added Meta as an AI sales channel, and on 21 September it announced Shop Pay checkout inside Muse (Yahoo Finance; Naughton & Bird). Amazon, notably, blocks it. Whatever you think of shopping by assistant, for Shopify brands it is no longer hypothetical.

Disclosure: we build TrueTag, a Shopify app for per-unit serials and verify pages. TrueTag has no connection to Meta or Muse. The argument below holds whichever tools you use.

What an agent actually sees

An agent shopping through Shopify's AI channels does not need to browse your store. It reads your catalog. Shopify's help page on agentic storefronts says Shopify Catalog sends products with their title, description, options, images, price, availability and "other key attributes", plus your store policies. Custom metafields only travel if you map them yourself, and merchants are shared by default (Shopify Help Center).

Now list what does not appear there: your homepage, the trust badges in your theme, the authorised retailers page, the video of your factory, the counterfeit warning you wrote after the third fake turned up on a marketplace. Most of the work brands have done to look trustworthy lives in places an agent never visits.

For brands in categories that get copied, such as electronics accessories, supplements, cosmetics and premium apparel, this matters twice. The agent comparing your listing with a lookalike sees two similar titles, two similar photos and two prices. Nothing in that comparison says which one can prove what it is.

Trust moves from the page to the product

If the storefront stops being where trust is built, it has to move somewhere an assistant and a buyer can both reach. There are really only two places that survive the trip: the product data itself, and the physical unit.

The product data.The description is the one field you control that reliably reaches an agent. A single factual sentence, such as "every unit ships with its own serial number and QR code; scan it to verify it is genuine", is now information an agent can weigh, the same way it weighs price and availability. Keep it plain and true. We would not promise that any particular agent repeats it. Nobody outside those companies can.

The unit.This is the part that cannot be copied from a listing. A code on the product that resolves to a public answer, genuine or not, which product, what warranty, works for the person holding the box. Increasingly it should work for the assistant that person asks, too. "I just got this, is it real?" is an easy question for an agent to hand to a public page, if that page gives a clear answer software can read.

That second half is the change worth making now. A verify page built only for human eyes forces the agent to scrape and guess. A verify result that also comes as structured data lets it report exactly what the brand published, and nothing more.

What to do this quarter

• Check what your catalog says.Read your top products' descriptions as if you were an agent with no access to your site. If authenticity matters in your category and nothing there mentions it, you are invisible on the one thing that separates you from a copy.
• Put one true sentence in the description. Only if every unit really ships with a scannable code. Do not claim what the product cannot back up.
• Make the proof live on the unit. Per-unit serials and QR labels mean the check happens after delivery, on the thing that was actually delivered, which is where counterfeits show up.
• Keep buyer data out of the answer. A public verify result should say what the product is, not who bought it. That is good privacy practice and it makes the result safe for any software to read.
• Do not wait for an integration. You do not need a partnership with an AI platform for any of this. A public, plain answer works wherever the buyer came from.

Where TrueTag fits

TrueTag gives every unit its own serial and QR label, assigned automatically as orders come in or created by you in bulk, with label printing free on every plan. A scan shows the item is genuine, the product and any warranty, and never buyer details. Each verify result is also published in a form software can read, so an assistant a buyer asks can check the same code. If you want the technical detail, our guide for AI agents explains exactly what a check returns. TrueTag helps detect and deter counterfeits. It cannot stop someone printing a fake box. What it does is give your genuine units an answer a copy cannot produce.

The free plan covers 50 units a month, enough to label one product line and see whether buyers, and their assistants, start asking.