Amazon flipped the switch last week. Rufus, the conversational AI assistant 300 million customers used in 2025, is now Alexa for Shopping, and it's rolling out across Amazon.com, the mobile app, and Echo devices right now.

Same conversational brain underneath, but with the household assistant fused on top and the assistant's memory now traveling with the shopper across every screen they own.

For sellers, this means your product is no longer being evaluated against a keyword. It's being evaluated against a persona Amazon has built over months and years of behavior.

Max’s team over at Azoma AI highlighted 8 changes to be aware of:

Eight Things That Are Actually Different

1. The main search bar is now a conversation. Shoppers don't need to open a separate Rufus window anymore. They can type a question — "what's a good skincare routine for oily skin in summer" — straight into the Amazon search bar. Amazon detects when the input is a question and routes it to Alexa for Shopping instead of the standard product grid. The standard PLP is no longer the default destination.

2. AI overviews now sit on top of search results and PDPs. A summary of the category, what to look for, and what matters most — generated by Alexa and placed at the top of search results and on product detail pages. The first thing a shopper sees about your category is now an AI-written summary pulling from listings, reviews, Q&A, and external content. Rolling out to all US shoppers now.

3. Side-by-side product comparison is native. Shoppers select multiple products from search results and Alexa lays them out against each other on features, price, and reviews. This used to live inside the chat. It's now part of the core search experience, and it's a direct counter to the "compare these five products" use case that's been pushing traffic to ChatGPT.

4. Scheduled Actions are agentic buying. Tap the "+" inside the chat and a shopper can set rules like "add this sunscreen to my cart when it drops to $10 and I haven't bought it in two months." Alexa runs the price check, does the research, and either notifies the shopper or adds the item directly. This is the first time Amazon has put true autonomous-buying behavior in front of every shopper, not just a Subscribe & Save segment.

5. Conversational cart rebuilds from purchase history. "Add my regular dog treats." "Add my favorite protein bars." Alexa pulls from the shopper's order history and builds the cart. One-tap checkout from there. Whoever the shopper said "my favorite" about wins.

6. Twelve months of price history on every PDP. A full year of pricing exposed with one tap, on hundreds of millions of products. The "fake sale" tactic of listing at $40, mark down to $24.99, and call it a deal is dead for any shopper who looks. Combine this with Alexa's price-tracking auto-buy and Amazon has effectively built Keepa into the native shopping experience.

7. Voice shopping is in the Amazon app. Tap the mic in the chat window and shop by voice from inside the Amazon app on a phone. You no longer need an Echo device to shop conversationally with Alexa.

8. The personalization layer is inspectable. Shoppers can ask Alexa what it knows about them. Their pets, family members, dietary needs and edits it on the fly. That's good news and bad news. Good because the data driving recommendations is no longer a black box. Bad because shoppers can now actively curate their profile in ways that will make some products less visible.

What This Means for Your Listings

The COSMO knowledge graph is still doing the semantic heavy lifting underneath. The retrieval logic is largely the same. If you've been optimizing for Rufus — clear noun phrases in titles, feature-to-benefit mapping in bullets, filled-out backend attributes, OCR-readable image callouts — that work isn't wasted. It's still the foundation.

But the bar is higher now, because three things changed at once:

Your listing is feeding more surfaces. The same content now shapes AI overviews on search and PDPs, drives side-by-side comparisons, and determines whether you get pulled into a custom shopping guide. Weak listings get glossed over in summaries or framed by a competitor's marketing language.

You're being matched to personas, not keywords. Alexa for Shopping doesn't just retrieve the most semantically relevant product for a query. It retrieves the most relevant product for this specific shopper, given their household, their history, their preferences.

Generic listings get sorted last. Listings that explicitly state who the product is for, how it's used, and what occasion or environment it fits get matched to the right shoppers.

Your competitors are visible to the shopper at the moment of decision. The side-by-side comparison surface is now native. If you win on features, price, and review quality, you get free placement next to your competitors.

If you lose on any one dimension, the AI is showing the shopper exactly why your competitor is the better choice.

What to Do This Week

OpenAI just opened the floodgates on ChatGPT advertising. Early data from Neil Patel shows affiliates already account for ~10% of ChatGPT ads on commercial terms. Translation: it's working, and the smart money is moving fast.

Where & How Ads Show Up

Sponsored links appear directly below ChatGPT's organic answers as visually separated callouts. Users can sometimes click through to keep the conversation going with the brand. No keyword matching. Ads are served contextually based on conversation intent, topic, and flow.

Who sees them: Free and Go plan users only. Plus, Pro, and Enterprise are ad-free. Free users can opt out entirely in exchange for fewer daily messages.

Where they don't show: Health, mental health, politics, and other regulated/sensitive topics are off-limits.

This is the gate most sellers will hit:

Campaign objectives (for now): Reach (CPM) or Clicks (CPC). Conversions coming — OpenAI's developer center already shows a pixel and Conversions API in the works, mirroring Meta's setup.

Targeting = "Context Hints": Plain-language description of who you want to reach and what queries should trigger your ad. Paste keywords, describe your ICP, describe the kinds of questions your customer asks. Combines keyword intent with audience targeting — closer to Meta + Google had a baby than either platform alone.

Ad format: Tight space. Concise headlines with numbers and clear benefits, strong CTAs ("Shop Now," "Get Discount"), and creative that grabs attention in a chat interface. Early winners follow this pattern.

Bulk editing: Currently clunky — export to Google Sheet, edit, re-import as CSV. You can also import Google Ads data this way as a shortcut.

Team & API: Invite team members, generate API keys for third-party tools.

Ads run on a system separate from the core chat model. Advertisers cannot influence, rank, or alter ChatGPT's actual answers, and they don't receive personal data or conversation transcripts for targeting. This matters because it shapes how users will perceive sponsored results long-term and likely keeps ad inventory premium.

If you're US-based with an EIN, get in the queue now. Early advertisers on every new platform (Google '02, Facebook '08, TikTok '20) captured outsized ROI before bidding got competitive.

If you're outside the US, watch Canada closely as the next domino, and start building your context hints + creative library now so you're ready to launch day one when your country opens.

The targeting model — intent + ICP description in plain language — rewards sellers who deeply understand their customer's actual questions. That's a structural advantage for anyone already doing voice-of-customer work.

Shopify just dropped data that should make every Amazon seller sit up straight. AI search isn't just sending traffic, it's sending better traffic than organic search, with higher conversion rates and higher order values. And the reason why has everything to do with how the buyer journey is changing.

Journey Compression Is the Whole Story

Here's the killer stat: 55% of AI-referred sessions start on a product detail page, versus just 20% for organic search. AI platforms recommend specific products, not categories or brands, so shoppers land already knowing what they want.

Think about the old journey: 3-5 searches, 8-12 pages across multiple sites, then maybe a return visit days later to actually buy. Discovery, consideration, and purchase were spread out across sessions.

Now compress all of that into one conversation. The shopper tells ChatGPT (or Copilot, Perplexity, Claude, Gemini) what they need. The AI narrows the options, surfaces a short list, and by the time the click happens, they're ready to buy.

That's "journey compression," and it's why per-session economics on AI traffic are already beating organic.

The Infrastructure Just Got Real

Every commerce wave needed a buildout before it tipped: web needed payment rails, mobile needed responsive design, social needed shoppable posts. AI commerce just got its equivalent:

Translation: the rails are live. The buying inside the chat — no click-through required — is coming fast.

The Historical Parallel You Need to Hear

In 2012, merchants who built for mobile before it tipped captured share that latecomers never got back. Same story in 2019 with social commerce.

We're sitting in that exact window right now for AI commerce.

What Sellers Should Actually Do

The Shopify piece outlines five moves. Here's the Amazon-seller translation:

Kevin's Take

This is the Amazon seller version of the mobile inflection point. Sellers who only think about Amazon SERPs are about to get blindsided, not by another Amazon algorithm tweak, but by the fact that buyers are skipping the Amazon search box entirely and asking an AI which protein powder, which dog harness, which kitchen gadget to buy.

That AI is making a recommendation based on signals from across the entire web, not just your Amazon listing.

If your brand has no presence outside Amazon, no editorial mentions, no Reddit conversation, no review coverage, no comparison content, you're invisible to the AI making the recommendation.

The window to fix that is open right now. It won't be open forever.

The new cart is an AI agent. Build for it.

✌🏼 See you again Thursday …

The answer to today’s STUMP BEZOS is
Amazon has 1.46 billion worldwide shoppers