Amazon's Shop Direct expansion is more than a feature update. It signals a structural shift in how product discovery and transactions will work. Here's what's changed, why Amazon has the AI commerce lead and what brands should do.
Amazon has announced a meaningful expansion of [Shop Direct](https://www.aboutamazon.com/news/retail/amazon-shop-direct-external-stores), its AI-powered shopping experience that surfaces products from external merchant sites inside Amazon search.
Merchants can now join more easily through third-party feed partners including Feedonomics, Salsify and CEDCommerce, with catalogue, pricing and inventory syncing in real time.
Customers can either click through to buy on the merchant's own site or, in some cases, use Amazon's "Buy For Me" flow, where Amazon completes the purchase on their behalf. Amazon says Shop Direct now includes more than 100 million products from over 400,000 merchants and is available to all US customers across Amazon.com, the shopping app, mobile web and Rufus.
> **Update — May 2026:** Amazon has merged Rufus with Alexa+ to create **Alexa for Shopping**, now live on the Amazon Shopping app, website and Echo Show. References to "Amazon Rufus" in this post relate to the predecessor product. [Read Amazon's announcement.](https://www.aboutamazon.com/news/retail/alexa-for-shopping-ai-assistant)
This is more than another feature update. It is Amazon making [agentic commerce](/blog/what-is-agentic-commerce-2026) operational rather than theoretical.
## What's actually changing
Two shifts are stacked inside one announcement.
**Amazon is no longer a closed marketplace.** It is becoming a discovery layer for the wider web. Product visibility is starting to depend less on traditional category ranking alone and more on whether your product data can be interpreted, matched and trusted by AI-driven systems. The same feeds merchants already use elsewhere can now power visibility in both standard search results and [Rufus](/blog/what-is-amazon-rufus-2026).
**Amazon is inserting itself into transactions that happen off-Amazon.** In the Shop Direct model, customers can be redirected to the merchant's store. In the Buy For Me model, Amazon acts on the customer's behalf using encrypted payment and delivery details, while the merchant still handles fulfilment, returns, exchanges and customer service. That is the agentic commerce protocol in production.
For challenger brands, this creates both opportunity and pressure. The opportunity: if Amazon refers high-intent shoppers to products not stocked in its own store, brand discovery expands beyond your existing retail footprint. The pressure: discoverability now depends on feed quality, data completeness and content written for machines as well as humans. If your titles, attributes, taxonomy and availability data are weak, AI-led surfaces struggle to represent you properly.
## Why Amazon's AI lead is structural
This isn't an isolated launch. It's the latest move in a position that's been building for years. AI at retail is a two-engine game - data to train on and compute to process it - and Amazon has both at world scale.
**Unmatched retail data to train on.** Independent estimates place around 2.3–2.8 billion monthly visits to Amazon.com in 2025. That's orders of magnitude more than most competitors. Walmart.com, by comparison, sees around 0.5 billion monthly visits. Eyeballs at that volume produce dense behavioural chains: query → view → cart → purchase → review. That's exactly what modern retail models train on. Tokens are the easy bit. The hard bit is behavioural data tied to structured PDP data and Amazon has the deepest, freshest chains.
The marketplace flywheel adds breadth. More than 60% of store sales come from third-party sellers, which means broader SKU coverage, richer attributes and faster content iteration than any single-retailer catalogue can match.
**Compute at scale, on tap.** AWS remains the leading cloud by market share, with specialised silicon (Trainium, Inferentia) plus partner GPUs. That's the substrate to train, fine-tune and serve retail models at global scale. AWS's lead is closely pursued by Microsoft and Google, but no other player combines hyperscale cloud with hyperscale retail traffic under one roof.
**The model is already inside the store.** Rufus is trained on Amazon's selection and blended with web information, embedded directly in the app and desktop buying flow. Assistants that live outside the cart must hand off to retailers. Rufus starts where intent becomes purchase, so it learns faster and converts smoother.
The combination - retail data plus compute plus an in-store assistant plus checkout - is the full stack that AI search and retail rewards. Today Amazon is the only player that controls all four. Walmart, eBay and others can partner with foundation-model providers, but stitching together external models, smaller behavioural graphs and fragmented infrastructure means slower iteration than Amazon's in-house flywheel.
## The early signals
This isn't theoretical. The behavioural data is already showing up.
According to Similarweb's 2025 eCommerce report, referrals from ChatGPT now convert at roughly **double the rate of organic search** and Amazon consistently leads among retailers capturing that traffic. AI referrals represent a small slice of overall traffic today, but a disproportionately high-intent slice. Shoppers entering through ChatGPT are further down the funnel, looking for a specific answer or product. They've already filtered, clarified, compared. By the time they arrive, they're validating, not exploring.
Amazon wins this traffic because it's built for structure. Complete data, clean taxonomy, deep reviews, real-time inventory. It is, functionally, the most machine-readable product graph on the internet.
So when an AI agent - ChatGPT, Perplexity, Gemini - needs a product answer, Amazon is often the safest place for the model to point. The trust is already pre-baked into the data layer.
For everyone else, the message is the same. Optimise for model visibility, not just search visibility. Because in AI commerce, the shelf space you don't see yet is the one that matters most.
## What brands should do now
This isn't a moment to panic. It's a moment to align.
**Make your catalogue answerable.** Every product attribute should be complete, accurate and aligned across Amazon, Shopify and structured feeds. Inconsistency is the silent killer.
**Publish agent-grade markup.** Treat Schema.org and product metadata as your visibility layer, not an SEO afterthought. Our [foundational guide to Schema.org for LLM optimisation](/blog/what-schemaorg-foundational-guide-llm-optimisation-2026) walks through the structured data that AI systems actually use.
**Optimise for inclusion, not ranking.** The goal isn't to appear first on Google. It is to be one of the few brands surfaced confidently in a chat interface. That's a different optimisation target.
**Track AI referrals early.** They're small now but carry the highest conversion rates of any channel emerging today. Set up Share of Model tracking across Rufus, ChatGPT, Gemini, Claude and Perplexity. Without a baseline, optimisation has nothing to compare against.
**Treat ads as ranking signals, not just traffic.** You can't buy every answer position, but you can create the behavioural proof that reinforces inclusion - strong CTR, saves, add-to-carts and review velocity on the exact queries you want to win. Sponsored Products as ranking-signal generators, not just demand harvesters.
## The takeaway
Commerce is moving from search engine optimisation to answer engine optimisation and now into agentic commerce optimisation. Brands need to think beyond webpages and PDPs as static assets. They need to treat product data as a live signal layer that feeds marketplaces, AI assistants, recommendation engines and autonomous buying flows.
If Amazon is opening discovery to external merchants through AI-powered experiences, brands should not treat this as just another Amazon feature update. It is a sign of where shopping is going.
The brands that win will be the ones easiest for AI systems to understand, trust and recommend.
That is the shift.
---
*If you want a baseline reading of where your brand stands inside Rufus and the broader AI commerce stack today, [Lmo7 can run a Share of Model audit](/contact) and a two-week media-to-rank pilot on your top SKUs.*