On 13 May 2026, Amazon quietly removed the single most important variable in Amazon advertising: the search box as we knew it. The Rufus chatbot was retired and replaced by Alexa for Shopping — an AI assistant that now sits in the main search bar, writes an answer above your organic results, and runs side-by-side product comparisons inside the results page itself. For sellers, this is not a cosmetic rebrand. It is a structural change to how products get discovered, and it quietly rewrites the rules of every PPC campaign you are running right now.
If your campaigns are still built the way they were in 2023 — exact-match keywords, manual bids, ACoS as the only KPI — you are optimising for a storefront that fewer and fewer shoppers actually use. This guide breaks down what changed, why it matters for your ad spend, and the exact campaign structure we are now deploying for the brands we manage.
Amazon launched Rufus, its generative-AI shopping assistant, to all US shoppers in 2025. By the time it reached scale it was serving an enormous share of shopping sessions, and on its Q3 2025 earnings call Amazon's leadership said the assistant was on track to contribute roughly $10 billion in additional annual revenue. Then, on 13 May 2026, Amazon retired the Rufus name and merged the technology into Alexa for Shopping, rolling it out as the default experience for every signed-in US customer — no Prime membership and no Echo device required.
The underlying recommendation logic and data sources stayed largely the same. What changed is prominence and placement. Instead of being a chat panel a curious shopper had to open, the AI now lives in the primary search experience: it interprets the query, generates an overview, and recommends specific products — including comparisons — before a shopper ever scrolls a traditional grid of results.
Independent holiday-season analysis suggested AI-assisted sessions accounted for roughly 40% of Black Friday traffic and around two-thirds of purchases during the 2025 peak. Whether or not those exact figures hold for your category, the direction is unmistakable: the AI layer is now where a large slice of buying decisions are formed.
Classic Amazon PPC rests on one assumption: a shopper types a keyword, sees a results page, and your ad competes for a slot on that page. Alexa for Shopping erodes that assumption in three ways.
First, intent replaces strings. The assistant interprets why someone is shopping — "a quiet blender I can use early in the morning without waking the baby" — and maps that to products through behavioural and contextual signals, not a literal keyword string. A bid on the exact phrase "blender" no longer guarantees relevance to that shopper.
Second, the funnel compresses. The AI summarises, compares, and shortlists in a single step. Consideration that used to happen across several page views now happens inside one answer. If your product is not in the consideration set the model assembles, you are invisible — regardless of how much you bid.
Third, the listing becomes ad infrastructure. The model reads your title, bullets, A+ content, and reviews to decide whether your product fits the shopper's intent. A weak listing now suppresses your reach before the auction, not just your conversion rate after the click.
The practical takeaway: PPC and listing quality are no longer separate disciplines. The strength of your value proposition — how clearly your detail page communicates who the product is for and what outcome it delivers — directly governs how often the AI is willing to recommend you. You cannot bid your way around a listing the model does not understand.
The most important advertising change is that Sponsored Products can now appear inside Alexa for Shopping results — within the AI's recommendations and comparisons, not only on the legacy keyword results page. For the first time, paid placements can surface during the AI-assisted discovery phase, at the exact moment a shopper is forming a shortlist.
Several industry analysts project that a meaningful portion of Sponsored Products budget — commonly estimated in the range of 15–25% — will flow toward these AI-driven placements over the course of 2026. Treat that figure as a planning signal rather than a guarantee, but plan for it: a growing share of your impressions will be served in contexts you cannot control with a keyword bid alone.
Here is the structure we are now rolling out across the accounts we manage. The organising principle is simple: build campaigns around buyer personas and use-cases, not isolated keywords.
One or two well-funded broad-match and auto campaigns whose job is not immediate ROAS but signal generation. These campaigns teach Amazon's matching engine which intents your product fits. Mine their search-term reports relentlessly — this is where you discover the conversational, intent-rich queries the AI is mapping to your product.
Exact and phrase campaigns built from the winning terms surfaced in Layer 1. These remain your efficiency engine. The difference in 2026 is that you are no longer guessing keywords — you are promoting terms the AI has already validated as relevant to real shoppers.
Branded defence (protecting your own terms in AI comparisons) plus targeted competitor conquest using Sponsored Products and Sponsored Display. When the AI runs a side-by-side comparison, you want to be the contender, not the absent option.
Negative keywords matter more, not less. As broad and auto campaigns expand to feed the AI, wasted spend can balloon. Disciplined negative-keyword and n-gram mining is what keeps your conversion funnel efficient while you widen the top of it. We treat this as a weekly ritual, not a quarterly clean-up.
Alongside the shopper-facing AI, Amazon has been rolling out an AI campaign-management layer — natural-language campaign creation, automated bidding, and creative generation handled by an "agent" rather than a human in the console. Used well, it removes hours of manual bid-tending. Used carelessly, it optimises toward whatever metric you point it at, including the wrong one.
Our rule of thumb: automate the execution, own the strategy. Let the agent manage intraday bid adjustments and budget pacing. Keep human control over campaign architecture, target ACoS/TaCoS by category, brand-defence rules, and the creative narrative. The agent is a fast junior trader; you are still the portfolio manager.
When a single AI answer blends organic and paid, brand and non-brand, discovery and conversion, attribution gets murky fast. This is exactly why we have argued for years that ACoS alone is misleading. In an AI-first world, lean on:
In the accounts we've restructured this way, ACoS and TaCoS both trend down over the following months as the discovery layer feeds organic rank — but TaCoS drops faster, which is the tell that paid spend is genuinely lifting organic discovery rather than just buying isolated sales.
QIs Amazon Rufus gone for good?
Yes. On 13 May 2026 Amazon retired the standalone Rufus chatbot and folded its technology into Alexa for Shopping, now the default AI assistant for signed-in US customers across the Amazon app, Amazon.com, and Echo Show. The recommendation logic is largely unchanged — the brand name and placement changed.
QDo Sponsored Products really appear inside the AI answers?
Yes. Sponsored Products can surface within Alexa for Shopping recommendations and comparisons, letting paid placements reach shoppers during AI-assisted discovery rather than only on a keyword results page.
QShould I stop using exact-match keyword campaigns?
No — keep them as your conversion engine. The shift is that you now also need well-funded discovery campaigns to feed Amazon's intent engine, plus listings strong enough for the AI to understand and recommend.
Sources & further reading: CNBC (Amazon ditches Rufus for Alexa shopping agent), GeekWire (Amazon unifies Alexa+ and Rufus), Tinuiti (Amazon agentic commerce), and Amalytix (Alexa for Shopping 2026 guide). Spend-shift and CTR ranges are industry projections and should be validated against your own account data.