As of 13 November 2026, Amazon's shopping assistant sits above the results page in the UK as well as in the US, and it now shows up to a full year of price history inside the answer itself. The company says more than 400 million customers worldwide used its AI shopping assistance in the past 12 months, and that the assistant is free for any signed-in customer, with no Prime membership and no Echo device required. For a pricing or ecommerce team, the operational change is narrower than the headline and more awkward: the reference price a shopper can check in one tap is no longer the one your promotion was built against.
What the UK launch changed and what it did not
The feature list is public and specific: overviews in search and on product pages, side-by-side comparisons of features, prices and review highlights, price history over three windows, personalised shopping guides for bigger purchases, and a basket built from past orders. What Amazon has not published is how the shortlist is ordered, how much of the catalogue a single answer can draw from, or what share of sessions end without a click on the results below. Those gaps matter because they decide whether the overview is a summary of the page or a replacement for it.
The scale claim is the company's own. On its third-quarter earnings call, chief executive Andy Jassy put the figure at more than 400 million users in the past year and said customers who use the assistant spend more per order than those who do not. He also played down competing agents, saying few customers do much shopping that way yet. Both statements come from the same call and neither is audited, which is why they belong in a planning note rather than in a forecast.
A year of price history changes the reference point
Price monitoring used to compare a shelf price with a competitor's shelf price on the same day. The assistant compares a price with its own past. A 30-day window captures the pre-promotion markup that most discount claims are built on; a 365-day window captures the season before, including the week when the same item was cleared out. A brand that plans a promotion against the short window can be technically honest and still look expensive against the long one.
Three windows, three different stories
The practical consequence is that a single discount percentage no longer carries the message. Teams that publish a reference price should decide in advance which window they are defending and make the comparison visible in the listing, because the shopper can now open the other two without leaving the page. That is a listing-content decision as much as a pricing one: the number needs a sentence next to it explaining what changed and when.
What money cannot buy in the answer
Amazon has not described a paid slot inside the assistant, and sponsored products remain a separate labelled unit on the page. That makes the answer an earned surface, shaped by catalogue data, review content and price rather than by bid. Third-party guidance converges on the same inputs the company itself describes: listing details, customer reviews and community questions, plus information from across the web. The ranking weights are not published, so any agency claiming a formula is guessing.
A price-and-listing check before the next promotion
The check below is written to be run per ASIN and per category, not per account. The middle column exists because the assistant reads structured data before it reads prose, and the right-hand column says what each number is for.
| Check | How to measure | What it decides |
|---|---|---|
| Reference price by window | Lowest price in the last 30, 60 and 365 days | Whether the discount survives the long view |
| Structured attributes | Share of required fields completed per ASIN | Whether the item can be matched to a use case |
| Negative review velocity | New one and two-star reviews per week | Whether sentiment is turning before the rating moves |
| Question coverage | Share of top buyer questions answered on the page | Whether the answer has material to quote |
The fourth check is the one teams skip. Answers are assembled from what the listing and the community section already say, so an unanswered question is a sentence the assistant cannot use. Filling it costs an afternoon and changes nothing about the price, which is precisely why it is the cheapest of the four and the last to be scheduled.
FAQ
Does the assistant replace the search results page?
No. Amazon describes an overview at the top of results and on product pages, with comparison on request. The list still exists underneath. What changed is that the first thing a shopper reads is a written recommendation with a shortlist, so items inside it get seen before anything below.
Why does a 12-month price history matter for promotions?
Because it changes the reference point a shopper can check without leaving the page. A discount measured against last month can look generous while the same item sat lower three months ago, and that comparison is now one tap from the buy button.
Can a brand pay to appear in the answer?
Amazon has not described a paid slot inside the assistant and says the shopping experience is free for signed-in customers. Sponsored products remain separate and labelled. Treat the answer as earned space shaped by catalogue data, reviews and price.
How reliable are the assistant's answers?
Amazon says accuracy is about 97% globally and that leaked internal figures are outdated. A document obtained by Reuters described answers as sometimes wrong or incomplete, with users unable to spot the errors. Both can be true: 97% still leaves roughly three answers in a hundred with a problem.
