As of November 7, 2026, Amazon has begun testing a way to shop by text message. The company said on November 6 that Alexa for Shopping can now take an order through a text thread, starting with an invitation-only group of US customers and expanding to millions of users over the coming weeks. The setup requires a verified mobile number and uses RCS messaging, and Amazon says the experience adds more than 100 new capabilities, including access to a year of price history. For brands, the interesting part is not the chat interface but the fact that a conversational order is assembled from the same catalog fields a product page already has, and those fields are frequently incomplete.
What Amazon actually shipped
Text-based ordering is a narrower change than it sounds, because the thread does not replace the store but replaces the search box for a specific kind of purchase. It suits the shopper who already knows roughly what they want and does not want to open an app to find it. Amazon has spent years pushing reordering into voice and dash buttons for exactly this reason, and the pattern repeats here. What is new is that the conversation can answer follow-up questions about price and availability inside the same thread, without sending the shopper to a page.
The reported context is worth separating from the announcement itself. Secondary coverage cites roughly 400 million customers using Alexa, agentic usage up more than eightfold since early 2026, and a claim that shoppers who use the assistant spend about 50% more. Those figures come from press reporting rather than Amazon's own release, and the spending claim describes a correlation rather than a causal lift. Treating it as causal would overstate what a text channel can deliver in its first year.
Why the catalog, not the chat, is the bottleneck
Price history turns into an answer field
Once an assistant can reference a year of price history, it can answer questions that used to require a spreadsheet. It can say whether the current price is a real discount, whether the item was cheaper last month, or whether a bundle beats buying separately. A listing that carries a promotional price without a clear reference price gives the assistant nothing to compare against, and the answer becomes a hedge. Retailers that maintain a clean reference price and a documented promotion calendar are simply easier to answer about.
Attributes decide whether an item can be ordered
A text order is short by design. The shopper says a size, a count or a flavor and expects the assistant to select the right variant from the parent listing. That only works when the variation is expressed in structured attributes rather than buried in a title or a description. Missing size, count or compatibility fields do not just hurt conversion, they make the item ineligible for the shortcut, and ineligible items lose the placement entirely.
Listing fields that decide whether you get picked
The practical question is which fields a conversational order actually depends on. Four groups cover most of the risk, and all four are already part of a mature listing on any marketplace. The difference is that a gap in any of them used to cost a click, and now it can cost the order before a page is ever shown.
| Field group | What the assistant needs | Common failure |
|---|---|---|
| Variation attributes | Size, count, flavor as structured values | Variant only in title text |
| Reference price | List price plus promotion window | Inflated list price with no end date |
| Availability | In-stock status by fulfillment node | Stock synced nightly |
| Compatibility | Model and device fitment | Compatibility only in the FAQ |
The price group is where most brands have the least discipline. A list price that drifts upward before a promotion is the pattern regulators in China and the EU have been flagging, and it is also the pattern that makes an assistant's price answer wrong. Amazon's own third-quarter advertising revenue reached $19.64 billion, up 11%, which means sponsored placements will keep competing for the same answer slots. A brand that cannot be answered about accurately is a poor candidate for a paid slot as well.
Four steps to make a listing text-shoppable
The test is small today, and that is the useful part of it. Problems found with a few thousand invited users are cheaper to fix than the same problems found after general availability. The work also compounds, because every field that becomes machine-readable improves filtering and comparison in the meantime.
Audit variation attributes on the top-selling parent listings, and convert anything that lives only in the title into a structured field. Start with size, count and flavor, because those are the three the assistant has to resolve to place an order. A parent listing with a clean variation set is easy to confirm in a single turn.
Document the promotion calendar with explicit start and end dates, and make sure the reference price matches what the catalog claims. A list price that moves before a promotion turns a simple price answer into a wrong one. The same discipline is what price monitoring teams already apply to channel pricing.
Check availability sync frequency, especially for grocery and replenishable categories where a nightly refresh is far too slow. A shopper who is told an item is available and then receives a substitution does not blame the catalog, they blame the brand. Fifteen-minute syncs are common on mature platforms and should be the baseline.
Write the three questions a shopper would ask by text, answer them from the listing alone, and fix whatever cannot be answered. If the listing cannot state the size, the price history or the fitment, the assistant will hedge or skip the item. This exercise surfaces gaps faster than any audit template.
There is a broader signal here for ecommerce teams that have nothing to do with Alexa. Conversational ordering rewards catalogs that are already machine-readable and punishes catalogs that rely on a human reading a long description. The same work improves filtering, comparison and search, so the investment is not specific to one channel. Amazon Now already covers more than 450 towns globally with units up roughly tenfold year over year, which suggests the fast-delivery catalog will keep growing and the assistant will keep reading from it.
FAQ
How does text-to-shop work?
Customers in an invite-only US test can message Alexa for Shopping, confirm details in the thread and place an order, using a verified phone number and RCS messaging.
Who can use it now?
A small invited group of US customers, with Amazon saying it will expand to millions of users over the coming weeks.
Why does the catalog matter more than the chat?
Because variation, price and availability answers are drawn from listing fields. Incomplete fields make an item hard to order and easy to skip.
What should brands fix first?
Variation attributes and reference price. Both are already required for a mature listing and both directly determine whether an assistant can answer accurately.
