Last updated on August 28, 2026 at 10:24 AM
Amazon remains the No. 1 retailer in Digital Commerce 360's ecommerce-sales ranking, yet it sits at No. 158 in the publisher's AI Commerce Rankings. That gap is not necessarily a weakness. It reflects a deliberate strategy to restrict third-party shopping agents and keep product discovery, recommendations, data, and transactions inside Amazon's own AI ecosystem. Google AI Mode is building the opposite model, where an assistant owns more of the discovery interface while merchants and booking partners complete the transaction. For retailers, the strategic question is becoming unavoidable: should you maximize visibility inside external AI systems, or protect more of the customer relationship inside your own environment?
| The Beecommercer Daily Buzz |
| Date: Friday, August 28, 2026 |
| Brought to you by: beecommercer.com |
Today's theme is clear: agentic commerce is creating a new distribution decision for every retailer.
Amazon is betting that its scale, customer habit, logistics network, advertising business, and first-party commerce data are strong enough to keep AI shopping inside its own walls. Other retailers are taking the opposite path, exposing catalogs and inventory to third-party assistants in exchange for more discovery.
Neither strategy is automatically right. The answer depends on how much demand the retailer already owns, how differentiated the catalog is, how portable the customer relationship needs to be, and whether the business can support reliable transactions when an external AI system sends a qualified shopper straight to the product or checkout.
Let's get into it.
THE BIG STORY
Amazon Is Deliberately Building A Walled AI Commerce Ecosystem
Digital Commerce 360's latest AI Commerce Rankings analysis shows how different Amazon's strategy is from retailers trying to maximize exposure through third-party AI systems.
Amazon remains No. 1 in Digital Commerce 360's ecommerce-sales ranking, but it sits at No. 158 in the AI Commerce Rankings. The AI ranking evaluates factors such as catalog accessibility to third-party agents, traffic from AI sources, source diversity, and 90-day momentum.
The low AI-commerce ranking is not necessarily evidence of failure. It reflects a deliberate business choice. Amazon has restricted external bots and has strong incentives to keep product discovery inside its own ecosystem, where Alexa for Shopping, Buy for Me, AWS, and Amazon's commerce data can support its own agentic strategy.
Walmart, by comparison, is No. 37 in the AI Commerce Rankings and is much more open to third-party AI discovery. Smaller retailers such as Nixon, Online Labels, and Everlane have ranked near the top despite sitting much lower in the traditional sales ranking.
This creates a real strategic choice for retailers. Open catalogs can increase discovery through ChatGPT, Gemini, and other external assistants. A more closed model can preserve control over product presentation, first-party data, recommendations, advertising economics, and the transaction itself.
Amazon can afford to choose the closed path because it already owns enormous customer demand and has the infrastructure to make its own AI shopping products useful. Most smaller retailers do not have that advantage. For them, refusing external AI access may protect control while also removing a growing source of discovery. The right strategy therefore depends on whether the brand is already a destination or still needs other platforms to introduce new customers.
| Operator takeaway: Decide explicitly which parts of the commerce journey you want external AI systems to access. If discovery is the constraint, make product data, price, availability, reviews, shipping, and return information easy for trusted agents to retrieve. If customer ownership is the moat, invest in an owned assistant, loyalty, first-party data, and a transaction experience strong enough that shoppers have a reason to begin inside your ecosystem. Do not drift into an open or closed model by accident. |
THE SHIFT
Google Is Building The Opposite Model: Open Agentic Commerce
Google has expanded AI Mode in Search with three new travel capabilities: flight-price tracking, points-and-miles comparisons, and direct hotel booking.
The travel category is a useful preview of what agentic commerce can look like because the decision combines structured data, personal preferences, dynamic prices, loyalty programs, reviews, cancellations, and payments.
Flight-price tracking is now available in more than 180 countries and territories. A user can describe the route and dates conversationally, ask AI Mode to track the fare, and receive an email when the price changes. Google says the underlying options use current data from more than 300 airline and travel-site partners.
AI Mode can also show the cost of flights and hotels in points or miles. Initial loyalty-program support includes Alaska Airlines / Hawaiian Airlines, American Airlines, Choice Hotels, Hilton, and Wyndham, with more programs expected to follow.
The most important change is hotel booking. In the United States, users can now move from a conversational hotel search to room selection, cancellation-policy review, and payment through Google Pay when an integrated partner supports the transaction.
Google is not becoming the merchant of record. The hotel chain or booking platform still handles the commercial relationship and customer service. Initial partners include Booking.com, Expedia, Hilton, Hotels.com, IHG, Marriott, Priceline, Trip.com, Wyndham, and others.
That separation is strategically important. AI Mode controls more of the decision interface, while the commerce partner remains responsible for inventory, price, payment acceptance, cancellations, refunds, and post-purchase support.
This may become a common agentic-commerce model beyond travel. AI owns the conversation and recommendation layer; the merchant supplies real-time commercial data and completes the transaction.
| Operator takeaway: Audit the steps between recommendation and transaction in your own category. Ask whether an external assistant could reliably retrieve current price, inventory, variants, shipping, return terms, loyalty benefits, and availability. If customers increasingly arrive after an AI system has already narrowed the options, the product page and checkout need to confirm the recommendation immediately rather than forcing the customer to restart the research. |
THE OPERATIONS LAYER
Agentic Commerce Still Ends In Customer Service
Shopify's new contact-center technology guide shows what has to happen after any AI-driven discovery channel succeeds: customer communication, order data, self-service, automation, workforce planning, and analytics all need to stay connected.
Shopify organizes the support stack into six layers: communication channels, helpdesk and CRM, knowledge base and self-service, automation and AI, workforce management, and analytics and quality assurance.
The strongest examples come from connecting support with real commerce data. Backyard Butchers reported a 70% increase in the speed of customer order changes and roughly a 50% reduction in ticket-resolution time after connecting support more tightly with Shopify. Tella & Stella reported 23% year-over-year sales growth after unifying DTC, B2B, POS, customer service, lifecycle marketing, and AI-assisted support on a more connected stack.
The implication for agentic commerce is important. Even when AI handles more of discovery and checkout, the merchant still owns the complicated part: the customer with a delivery exception, refund request, wrong variant, damaged item, or policy question.
| Operator takeaway: Treat customer service as part of the agentic-commerce architecture. Connect support to live order, product, customer, and policy data so an AI-generated transaction can still be changed, refunded, explained, or corrected accurately. The more automated discovery and checkout become, the more valuable a reliable exception-handling system becomes. |
3 THINGS IN ADS, CREATIVE & PLATFORM OPERATIONS
1. Demand Gen Pushes The Open Model Closer To The Transaction
Google's August Demand Gen Drop adds three notable capabilities: direct messaging entry points from YouTube ads, more travel-specific inventory and offers, and generally available multimodal video creation inside Asset Studio.
Google says the hundreds of Demand Gen improvements introduced during the second half of 2025 drove an average 30% increase in conversions or conversion value for advertisers. That is platform-reported experimental data, not a guarantee for every account, but it explains why Google is pushing the campaign type further into lower-funnel use cases.
The messaging test matters because the next action can now be a conversation instead of a website visit. A customer watching a YouTube ad may be able to open a messaging app and talk to the brand while intent is still high.
Travel advertisers will also be able to surface local activities, events, real-time offers, and more personalized hotel advertising. The theme is the same as AI Mode: recommendation is becoming more context-aware and closer to the transaction.
The creative layer is changing at the same time. Multimodal Video Creation in Asset Studio is now generally available, allowing advertisers to storyboard and create horizontal and vertical assets in a single workflow.
Google's WAAW by Alok case study shows the level of governance required when personalization becomes generative rather than template-based.
The project started with a fully human-authored official music video and then used Gemini Omni, Maps, Weather, YouTube & Trends, Vertex AI, and Veo to adapt approved parts of the experience to a fan's location, weather, style, and mood.
The interesting number is not how many versions the system could generate. It is how many the team rejected. Of 23,234 candidate outputs, only 4,320 – 18.6% – passed the project's Artistic DNA Gate. The rest were rejected or sent back for refinement.
The lesson is that scalable personalization needs more review logic, not less. The team defined what could change, what had to remain fixed, and which visual and emotional criteria every output had to satisfy.
| Operator takeaway: Before enabling more generative creative, define the non-negotiables: product shape, packaging, claims, brand colors, tone, creator rights, and any scene elements that must remain fixed. Use AI to multiply approved ideas, but build review and rejection into the workflow. For messaging-based ads, make sure the customer reaches a team or automation system that can answer the exact question the ad is likely to create. |
2. Sponsored Results May Become More Interactive
Google is again testing swipeable categories inside sponsored results. Users can swipe among category labels that alter the sponsored results shown below, with examples such as a featured category and a cheaper-price category.
The experiment resembles a small layer of guided shopping inside the ad unit. Instead of treating every commercial query as one fixed results list, Google can let the user refine intent within the sponsored experience.
For ecommerce advertisers, this makes product classification, price positioning, and feed quality more important because the platform may increasingly reorganize sponsored inventory around category-level intent.
| Operator takeaway: Do not optimize around an unconfirmed test, but monitor live Shopping and Search results in your highest-value categories. Keep product taxonomy, titles, pricing, and promotional data clean enough that Google can place the right SKU when users refine their intent inside the ad experience. |
3. Google Is Publishing More Structured Guidance For Performance Fluctuations
Search Engine Roundtable reports that Google has published or refreshed dedicated troubleshooting documentation for Google Ads, Performance Max, Search, Display, Video, Shopping, App, Smart, and Demand Gen campaigns.
The exact publication history of every help page is difficult to verify, but the direction is useful: Google is trying to give advertisers campaign-type-specific frameworks for diagnosing changes rather than reacting immediately to one metric.
That matters as bidding, creative, targeting, and placements become more automated. A performance decline can come from budget changes, conversion tracking, competition, seasonality, assets, feeds, policy, learning, or auction dynamics.
| Operator takeaway: Use a consistent diagnosis sequence before changing bids or budgets. Check conversion tracking, change history, budget status, auction conditions, product-feed health, creative changes, seasonality, and landing-page performance. Automation increases the need for disciplined troubleshooting because more variables can move outside the manual settings you changed. |
OFF THE CLOCK
Owning The Interface Also Means Owning The Responsibility
Meta Is Turning A Legal Settlement Into A Platform-Safety Campaign
Business Insider reports that Meta began running paid ads calling on TikTok and YouTube to adopt similar teen-safety measures after Meta agreed to a settlement of up to $18 billion in a teen-addiction lawsuit.
Meta's ads ask rival platforms to 'join us in supporting teens' and link to the company's explanation of the settlement protections.
Business Insider describes the strategy as reputational jujitsu: Meta is trying to turn a major legal settlement into an industry-wide argument about shared standards rather than allowing the story to remain focused only on Meta.
For advertisers, the important point is broader than the PR tactic. Safety, age assurance, product design, moderation, and youth-targeting rules can affect how social platforms operate and how brands are allowed to reach younger audiences.
| Operator takeaway: Treat platform safety and regulatory change as part of media planning, not a separate legal topic. In campaigns involving younger audiences, document age targeting, creative suitability, creator relationships, data use, and platform-specific restrictions. A large audience is valuable only while the platform can continue offering that audience under acceptable legal and trust conditions. |
Google's latest Workspace with Gemini update is aimed at students, but it reflects the same product direction visible across travel, ads, and commerce: AI is moving inside the tools where the task already happens.
Instead of opening a separate AI product, users can organize assignments, turn notes into study guides, draft presentations, summarize information, and manage work inside Workspace.
That product pattern is becoming common. Search is becoming a booking workflow. Demand Gen can become a conversation workflow. Contact centers are becoming AI-assisted resolution workflows. Amazon is building shopping agents inside its own ecosystem.
This is the strategic tradeoff behind Amazon's approach. The company is not simply blocking bots; it is protecting the interface where customer intent, product discovery, advertising, transaction data, and loyalty can reinforce one another.
A closed ecosystem can create stronger first-party economics, but it also requires the retailer to build enough discovery, recommendation quality, service, and trust that customers are willing to stay inside it.
An open ecosystem can gain reach through external assistants and search platforms, but more of the decision may happen outside the brand's direct control. Product data has to be accurate, partner permissions have to be intentional, and attribution becomes harder.
The winning model will not be the most open or the most closed. It will be the one that matches the retailer's real advantage: owned demand, unique inventory, logistics, service, loyalty, or the ability to become the most useful source wherever the customer begins.
The Beecommercer Team