Amazon Killed Rufus. The Real Story Isn't About Amazon.
Every brand I talk to this month is asking some version of the same question, and most of them are asking it about the wrong company. They want to know what to do about ChatGPT, Perplexity, and Gemini. Meanwhile Amazon quietly rewrote the rules of its own search bar.
"If the biggest store on earth just turned its search box into an agent, what does 'ranking' even mean anymore?"
TL;DR
Amazon is discontinuing the standalone Rufus chatbot and replacing it with Alexa for Shopping, a personalized agent that lives inside search results.
The search bar is becoming a Q&A engine. Browsing for a product now opens a chat window with recommendations, comparisons, and price triggers.
OpenAI already pivoted away from Instant Checkout toward dedicated retailer apps. Amazon is going the opposite direction: own the agent, own the customer, own the catalog.
For DTC brands, this means two front doors are being rebuilt at once. The external one (ChatGPT, Perplexity, Gemini) and the internal one (Amazon, Walmart, retailer agents).
The brands that win the next 18 months treat agent visibility as a discipline, not a campaign. They show up in the answer, not just on the page.
This is not an Amazon story. It is a discovery story. The question every founder should be asking is what "shelf space" means when the shelf is a conversation.
What's Actually Changing
For two years Rufus sat on the side of the Amazon experience as a curiosity. A chat tab. A beta. Most shoppers ignored it, and most brands treated it as a problem for next year. Amazon just made it the main event by folding Rufus and Alexa+ into a single agent called Alexa for Shopping, summoned by a cursive A icon on the site, the app, and Echo Show devices.
The mechanical change is small. The behavioral change is enormous. When a shopper browses for "running shoes for flat feet," Amazon now opens a chat window inside the results with a few recommended items and contextual information, drawing on the shopper's history, reviews, and Amazon's catalog.
It can compare products side by side. It can schedule a purchase when a price hits a threshold. Daniel Rausch, Amazon's top Alexa executive, framed the moat clearly: this is not scraping web results, this is acting on inventory data, delivery estimates, and review depth Amazon already owns.
Notice what Amazon is doing while OpenAI walked back Instant Checkout in favor of dedicated retailer apps inside ChatGPT. The two largest forces in AI commerce are moving in opposite directions. OpenAI is opening up a platform for retailers to build inside. Amazon is closing the loop and building the agent itself, while continuing to block many external shopping bots from accessing its site.
The real story is this: discovery is being rebuilt simultaneously on two surfaces, the open web answer engines and the retailer-owned agents, and most DTC brands have a strategy for neither.
What This Looks Like in the Wild
We have been watching brands react to this in real time. Three patterns keep showing up, and each one tells you where the visibility gap actually lives.
1. The Brand That Won SEO and Lost the Answer
A skincare brand we work with has page-one Google rankings for almost every category term they care about. They have a decade of SEO investment behind them, and the traffic numbers still look healthy on the dashboard. When we ran the same queries through ChatGPT and Perplexity, they were cited in roughly one out of every six relevant answers.
Their direct competitor, a younger brand with a fraction of the domain authority, showed up in more than half. The difference was not authority. The difference was the off-site content footprint: review depth, third-party roundups, structured comparison pages.
2. The Brand That Owns One SKU and Nothing Else
A wellness brand we know is famous for a single hero SKU. Inside ChatGPT, that product gets named almost every time someone asks for the category. Inside Amazon's new Alexa for Shopping experience, the same hero SKU surfaces in the recommended panel. The problem is everything else in their catalog.
The agent does not know they make a complementary product, a starter bundle, or a subscription option, because none of that content exists in the structured places the agent can read.
3. The Brand That Is Invisible in Both Worlds
A home goods brand with great product and a loyal customer base barely shows up in either surface. Not in ChatGPT answers. Not in Perplexity citations. Not in Amazon's new chat panel, even though they sell on Amazon.
Their reviews are sparse, their PDPs are written for humans only, and their off-Amazon presence is limited to their own site. The agent has nothing to grab onto. They have great product and zero shelf space in the new discovery layer.
The pattern across all three: traditional metrics tell you nothing about whether the agent knows you exist.
The Framework: The Two-Front-Door Audit
The mistake most brands are making right now is auditing one surface and ignoring the other. ChatGPT visibility matters. Amazon agent visibility matters. They draw on different signals, and you need to know where you stand on both before you spend a dollar fixing anything.
Here is the shape of the audit.
Front Door One: The Open Web Answer Engines
Pick ten shopper-voice queries that cover TOFU (top of funnel), MOFU (middle of funnel), and BOFU (bottom of funnel) for your category. Run each one in ChatGPT, Perplexity, and Google AI Mode. Examples to start from:
"What are the best [category] brands for [specific use case]?"
"Is [your brand] better than [competitor]?"
"What do reviewers say about [your hero SKU]?"
"I have [problem]. What should I buy?"
For each query, record three things: were you cited, what was cited about you, and what was cited about your top competitor. This is your external GEO (Generative Engine Optimization) baseline.
Front Door Two: The Retailer Agents
If you sell on Amazon, run the same queries inside Alexa for Shopping. If you sell on Walmart, run them through Walmart's emerging agent surfaces. Pay attention to whether the agent surfaces your hero SKU only, your full catalog, or nothing at all. Pay attention to whether the agent's summary of your product matches what you actually want shoppers to know.
The Spread Check
Plot your citations across both front doors. The brands that win are not the ones who dominate one surface. The brands that win show up consistently across both, with consistent product narrative in each. If your story changes between ChatGPT and Amazon Alexa, the agent picks the version with more structured evidence behind it, and that may not be the version you want.
Run It: Practical Checklist
Here is what to actually do this week. Not next quarter, this week.
Build the 10-prompt set and run it on three platforms. Write ten queries in the voice of your real customer. Run them in ChatGPT, Perplexity, and Alexa for Shopping (or another retailer agent where you sell). Save the outputs. Note every citation and every recommendation. This is your benchmark.
Run the catalog spread check. Inside each surface, ask the agent about three different products in your line, not just your hero SKU. If only the hero shows up, your structured content footprint is concentrated in one place and your second and third products are invisible. That is a content problem, not a product problem.
Audit your review depth where the agent looks. Amazon's agent leans on customer reviews and the depth of its product catalog. ChatGPT and Perplexity lean on off-site mentions, third-party roundups, and structured comparison content. Pull a count of reviews on your top five SKUs on every retailer surface you sell on, and pull a count of off-site editorial mentions for the same SKUs. Where reviews or mentions are thin, that is where the agent goes quiet about you.
Fix structured content where the agent reads it. Rewrite your top five PDPs (product detail pages) so the answer to "who is this for, what problem does it solve, how does it compare" lives in the first paragraph. Add FAQ blocks that mirror the shopper-voice queries from step one. Make sure category and comparison content exists on your owned site and in at least three credible third-party publications.
Set a monthly cadence. Agent visibility moves. ChatGPT updates its model. Perplexity changes its ranking weights. Amazon's Alexa for Shopping will iterate. Pick the first Monday of every month, rerun the ten-prompt set across both front doors, and track the delta. Treat it the way you treat your paid acquisition dashboard.
In closing...
What Amazon did with Rufus is the loudest version of a quiet shift that has been building for a year. The search box, on the open web and inside the largest retailers, is becoming a conversation. The brands that show up in that conversation get the click, the cart, and the loyalty. The brands that do not show up disappear from the dataset that the next agent will train on.
This is the front door of ecomm being rebuilt in front of us, and it is happening on two surfaces at once. Treating it as one problem will leave you exposed on the other. Treating it as a future problem will leave you exposed on both.
The work for the next 18 months is plain. Make your brand discovered, trusted, and chosen, in every answer, on every surface, by every agent.
Hope this was helpful. If it was, feel free to share it with a friend.
– Tomer
If you haven't assessed how your brand shows up in AI-powered discovery, the tool we launched covers that: https://commerce-gpt.yotpo.com
