What voice commerce is

Voice commerce covers purchases in which the shopper talks to an assistant instead of browsing a catalogue: “order groceries”, “repeat my last order”, “where is my parcel?”. The assistant recognises the speech, works out the intent and carries out the action through the store’s catalogue and order system. It is a subset of conversational commerce: the interface is speech, and there is no shelf.

How a voice purchase works

Stage What the assistant does What the store has to provide
Recognition Converts speech to text, picks out the product, quantity and brand Product names that are easy to say out loud
Selection Matches the request against the catalogue and purchase history Unambiguous attributes: volume, weight, flavour, size
Confirmation Reads the choice back or sends the basket to a screen Current price and availability
Order and service Creates the order, reports delivery status An API for orders and order status

Where it already works

  • Alexa+. In December 2025 Amazon added delivery tracking, reminders to reorder household essentials and the option to add items to an upcoming delivery to Alexa+ in the US and Canada. Price tracking and automatic purchases at a target price had arrived earlier.
  • Alice and Yandex Lavka. Yandex — the dominant search engine in Russia and several CIS markets — has let shoppers build a basket in Lavka, its rapid grocery delivery service, by voice since September 2023. The shopper lists products; if no brand is named, the assistant Alice picks one based on purchase history, availability and favourites. The finished order arrives as a push notification in the app, where it is edited and paid for.
  • Alice and Yandex Market. Through a Yandex Station smart speaker, a TV or the app, Alice reports the status of a Yandex Market order and when it will arrive.

For the Russian market the next step is Yandex Commerce Protocol (YCP), announced on 27 February 2026, which connects store catalogues to AI scenarios in Alice and Yandex Search.

Why the first position wins in voice

On a screen the shopper sees dozens of products and chooses for themselves. By voice the assistant reads out one or two options — listening to more is tedious. So the competition is not for the first page of results but for the first mention. The winner is a product the assistant can match unambiguously to the request (“whole milk, one litre”) and that is definitely in stock. Answer engine optimisation follows the same logic: there is one answer, and structured data is what earns it.

What a store should do

  • Check product names: they should read aloud with no SKU codes or abbreviations — “Pasteurised whole milk 3.2%, 930 ml”, not “MLK PAST 3.2 0.93”.
  • Fill in the attributes shoppers use to refine a spoken request: volume, weight, flavour, size, pack count.
  • Keep price and availability current — the assistant must never offer something that is out of stock.
  • Make reordering easy: purchase history and “my usual” are the main voice scenario.
  • Prepare your data for agentic channels — see agent-ready product feed and AI shopping agent.