As of 1 November 2026, Amazon has put a number on conversational shopping. On its fiscal third quarter earnings call, the company said more than 400 million customers have used Alexa for Shopping in the past 12 months, up from 350 million at the end of July, and that customers who use it spend 50 per cent more than those who do not. Chief executive Andy Jassy framed the assistant as the primary shopping agent customers will use to shop Amazon, even as he acknowledged the company expects to work with horizontal agents run by others.
What the 400 million figure actually measures
The headline number sits inside a quarter where retail growth was softer than the total. Net sales reached 201.9 billion dollars, up 12 per cent, but the bulk of that came from AWS, where segment sales rose 39 per cent. Online stores revenue grew 2 per cent, against 10 per cent in the same quarter a year earlier, and third party seller services grew 9 per cent against 12 per cent. Part of the gap comes from timing, because Prime Day landed at the end of June this year rather than in mid-July, moving that sales growth into the second quarter.
The public history of the assistant runs back three years. Amazon released an AI shopping assistant called Rufus in February 2024, rebranded it to Alexa for Shopping in May of this year, and has since made it more visible on the website and app. New capabilities added ahead of the holidays include the ability to create and edit gift lists and registries through queries such as showing what is left on a wedding registry, plus the ability to pause or cancel Subscribe and Save subscriptions from inside the assistant.
Why the entry point matters more than the growth rate
The registry is a data asset, not a feature
A gift registry is one of the few places where a shopper states intent months in advance and in public. Letting customers build and edit registries inside the assistant gives Amazon a structured list of items, quantities and dates that a search box never produces. For a brand selling into that list, the question stops being whether the product page ranks and becomes whether the item is present in the registry at all, which is a different kind of listing problem.
Vendors inherit a new visibility rule
When the assistant answers a shopping question directly, the product page is no longer the first thing a customer sees. Sellers who built their presence around keyword placement on the marketplace now depend on whether their items can be described, compared and recommended in a sentence. That shifts weight toward structured item data, consistent naming and accurate attributes, because an assistant that cannot tell two similar items apart will not confidently recommend either. Seen through consumer demand rather than page traffic, category growth now depends on how cleanly each item can be described.
Returns become the next battleground
Higher basket values from assisted shopping make returns more expensive to absorb. Marketplaces are already reworking return policies and reverse logistics as order volume grows through social and delivery channels, and DoorDash has been expanding its retail roster alongside a text based shopping assistant that lets customers order inside a messaging thread. Every one of those channels inherits the same question: who pays for the return when the recommendation came from a machine.
Channel comparison for assisted shopping
| Channel | Entry point | What the shopper sees first | Seller dependency |
|---|---|---|---|
| Marketplace search | Keyword box | A ranked list of product pages | Keyword placement and reviews |
| Alexa for Shopping | Spoken or typed question | A short answer with items | Structured attributes and item clarity |
| Messaging assistant | Chat thread | A suggested basket | Availability and delivery promise |
| Social marketplace | Feed and live video | Creator context and offer | Content cadence and return policy |
The second row is the one reshaping seller operations. In a ranked list, a shopper compares pages and the page does the persuading. In an answer, the assistant does the persuading and the page only confirms. That is why item level data quality now behaves like a distribution problem: if the attributes are incomplete, the item effectively does not exist inside the answer even though it exists in the catalogue.
What sellers should fix before the holiday peak
The first fix is attribute completeness on the top selling items, measured as the share of items with every field an assistant would need to describe them. The second is naming consistency, because two spellings of the same product line split the evidence an assistant uses to recognise the item. The third is the return promise, since an assisted purchase that arrives wrong costs more than a search driven one and the reverse logistics cost lands on the seller either way.
There is also a measurement problem. If a growing share of orders start with a question rather than a keyword, the attribution window a seller relies on will increasingly show direct or branded traffic instead of the actual entry point. The practical response is to tag assisted sessions where possible and to track item level demand alongside page level traffic, so that a shift in the entry point does not read as a sudden change in customer loyalty.
FAQ
How many customers actually use Alexa for Shopping?
Amazon said more than 400 million customers used it in the past 12 months, up from 350 million at the end of July. Chief executive Andy Jassy also said those customers spend 50 per cent more than shoppers who do not use the assistant, which is the figure that matters for basket economics.
Why did online stores growth slow in the same quarter?
Mostly because of timing. Prime Day fell at the end of June this year rather than in mid-July, so that sales growth was booked in the second quarter. Online stores revenue still grew 2 per cent, against 10 per cent a year earlier, while third party seller services grew 9 per cent against 12 per cent.
Does a gift registry inside the assistant change listing strategy?
It changes where the listing has to appear. A registry is a stated intent list built months ahead, so the question becomes whether an item is present in that list rather than whether a product page ranks for a keyword. That rewards structured item data over page level optimisation.
What should a seller measure first?
Attribute completeness on top selling items, because an assistant that cannot describe an item confidently will not recommend it. Naming consistency comes second, since duplicate records split the evidence, and the return promise comes third because assisted purchases that arrive wrong cost more to reverse.
