Ecommerce Weekly | AI carts go live as platforms start ordering and discounting on your behalf
October 12–18, 2026: Shop MCP, Google Universal Cart, Google Ads automated promotions, Amazon peak fees and returns, Shopify batch fulfillment
This week AI shopping moved from interfaces to products buyers can use. Shopify launched Shop MCP, letting any agent search Shopify Catalog, build carts, and check out with Shop Pay; Google launched Universal Cart in US Search and the Gemini app. Platforms also began changing operating settings without merchant action: Google Ads switched on automated promotions by default for some campaigns, Amazon's holiday peak fulfillment fees started on October 15, and Amazon made its January 31 holiday return deadline an annual rule. Shopify added batch picking, packing, and automatic label purchase. Each item is checked against official material. With Black Friday (November 27) less than six weeks away, actions are limited to what can be finished and read back before the sale.
Editorial linePublicThe week's main story: agents can place orders, and platforms start deciding for you
Open the editorial line and its boundaries
On October 13, Shopify launched Shop MCP: one MCP connection to search Shopify Catalog's more than one billion products, build carts, and complete Shop Pay checkout once the shopper confirms the total and payment method. On October 15, Google launched Universal Cart in US Search and the Gemini app, letting shoppers keep items from several merchants in one cart and pay with Google Pay or transfer to the merchant's site. On October 16, Shopify's Checkout MCP and Cart MCP added the UCP buyer-consent extension, so agents can pass marketing and privacy choices.
In the same week, another group of changes took effect without merchant action. From October 12, Google Ads switched on automated promotions by default for Search and Performance Max campaigns with location assets and no promotion assets. Amazon's holiday peak fulfillment fees started on October 15, based on ship date and covering FBA, MCF, and Buy with Prime. On October 16, Amazon made "buy in November–December, return by January 31" an annual rule.
For an independent store, both groups point the same way: prices, offers, returns, and consent settings are now read and executed by agents and platforms directly. Confirm those settings are what you intend before expanding AI channels.
Verified facts
- October 13: Shop MCP launched. Agents can search Shopify Catalog, build carts, and pay with Shop Pay; purchases complete only after the shopper confirms total and payment method; no webhooks at launch.
- October 15: Google Universal Cart launched in US Search and the Gemini app, with Google Pay checkout or transfer to the merchant cart; merchants remain seller of record and handle confirmations and support.
- October 16: Checkout MCP and Cart MCP support the UCP buyer-consent extension; without agent-sent choices, the shop's defaults for the buyer's country apply, and unknown countries default to opt-out.
- October 12: Google Ads turned on automated promotions by default for eligible campaigns, extracting offers from your site into ads; it can be switched off in account-level automated assets.
- October 15, 2026 – January 14, 2027: Amazon peak fulfillment fees average about $0.32 per unit across FBA, MCF, and Buy with Prime (trade reports of Amazon's July notice).
- October 15–16: Shopify added batch fulfillment (up to 250 per batch), automatic label purchase in Flow, Royal Mail pickup points in the UK, and DHL eCommerce labels in the Netherlands and Poland.
Black Friday falls on November 27, less than six weeks away. This issue keeps only actions that can be finished and read back before the sale.
Impact read
Macro view: the execution layer of AI shopping largely arrived this week. Agents can search, add to cart, and check out; platforms can add promotions and extend return windows. The data layer is still missing: how agent orders are identified, how transferred orders are attributed, what automated promotions delivered. So the focus now is not connecting more but confirming that every setting platforms read is accurate, and keeping definitions ready for the post-sale review.
1. Shop MCP launches: any agent can search, add to cart, and check out with Shop Pay
Shopify's October 13 developer changelog introduced Shop MCP. Through one MCP connection (https://mcp.shop.com), an agent or app can search Shopify Catalog's more than one billion products, build carts, and complete Shop Pay checkout. Main tools include search_catalog, lookup_catalog, get_product, complete_checkout, get_checkout, and search_orders. Carts and checkout work without sign-in; when the shopper signs in with Shop, search uses their preferences and checkout can use saved addresses and Shop Pay. The agent completes a purchase only after the shopper confirms the total and payment method, and order lookup requires sign-in. There are no webhooks at launch, so developers poll checkout status; agent-placed orders appear in the merchant admin like any other order.
On October 16, Shopify said Checkout MCP and Cart MCP support the UCP buyer-consent extension (from UCP 2026-08-25). Buyers can make the same privacy and marketing choices in an agent as in the store's checkout: analytics, preferences, sale or sharing, and marketing including email and SMS. If the agent sends no choices, the store's defaults for the buyer's country apply; if the country is unknown, consent defaults to opt-out. An SMS opt-in without a phone number blocks checkout until the agent supplies it.
What this means
Impact read
The agent ordering path is now complete: search, cart, checkout, and order lookup, with buyer confirmation and consent written into the protocol. Orders arrive like normal orders, but which agent sent them and what description the buyer saw happen outside your store.
What it means for an independent brand
Whether agents find your products and describe them accurately depends on your Shopify Catalog data: titles, attributes, prices, compare-at prices, inventory, and policies. Marketing consent can now arrive from agents, and unknown countries default to opt-out, which affects the size of post-peak remarketing lists.
Decision: the feature is certain and the path into the store is open, but demand and order identification remain unclear. Manage product data and consent defaults; do not chase agent traffic on its own.
Not yet
Avoid this week
- Do not rewrite product titles and descriptions in bulk before the peak for agents; check factual accuracy first.
- Do not mix agent-sourced marketing consent with on-site consent until you have confirmed the source field.
Decision boundary
Applies to Shopify merchants whose products are in Shopify Catalog. Impact: medium. Confidence: high on the feature, low on demand. The judgment changes if Shopify adds source markers and reports for agent orders, or if major AI assistants connect Shop MCP by default and produce visible orders. It proves any agent can order from your store; it does not prove incremental sales.
The smallest step this week: ask an MCP-capable AI assistant to find five of your lead products and record whether title, price, discount, and stock match the store; also check marketing-consent defaults for your main markets.
2. Google Universal Cart launches in the US: a cross-merchant cart that checks out via Google Pay or the merchant
Google's official blog announced on October 15 that Universal Cart launched in the US in Google Search and the Gemini app. Shoppers can keep products from several retailers in one cart and pick up where they left off across Search and Gemini. Google monitors price and stock for saved items and sends price-drop and restock alerts, uses Google Wallet to suggest which card to use, and surfaces relevant promo codes. Checkout runs through Google Pay or by transferring items to the merchant's own cart. Google says its Shopping Graph holds more than 60 billion product listings, names Nike, Sephora, Target, Ulta Beauty, Walmart, Wayfair, and "many Shopify merchants", and says merchants remain the seller of record and handle confirmations and support.
PPC Land reported the same day that YouTube and Gmail, mentioned at the May I/O preview, were absent from this launch, and that the post does not describe merchant fees, ranking rules, promo-code sourcing, or which cart data is shared with merchants.
What this means
Impact read
Comparing prices, waiting for markdowns, and hunting for codes used to happen across many tabs; now they sit in Google's cart. Price and stock changes are pushed to shoppers, so your price moves are seen faster and compared more easily.
What it means for an independent brand
For US-market independent stores, products enter the cart from Shopping Graph data, so accurate price, stock, and promo codes decide whether shoppers come back to buy. Promo codes are surfaced automatically, which means creator-specific or internal test codes may also be seen.
Decision: the product is live and the direction is clear; the direct effect on independent stores is more price transparency and promo-code leakage risk, while conversion impact is not visible yet.
Not yet
Avoid this week
- Do not set Universal Cart as a separate channel KPI until attribution of transferred orders is clear.
- Do not make frequent small price changes during the sale; price-drop alerts amplify volatility.
Decision boundary
Applies to independent stores selling in the US with products in Google Shopping. Impact: medium. Confidence: high on the feature, low to medium on commercial effect. The judgment changes if Google publishes attribution and data-sharing rules for transferred orders, or if Merchant Center reports show Universal Cart clicks and orders separately.
Google has not said when coverage will widen or reach other countries; we keep that on the watchlist. Micro action: list every promo code that works publicly and add conditions to, or disable, codes that should not be public; sample 20 lead products in Merchant Center and confirm price and stock match the store.
3. Google Ads automated promotions default on from October 12: Google pulls offers from your site into ads
On October 5, Search Engine Roundtable published Google's email to advertisers: from October 12, 2026, Google Ads "will automatically identify and extract high-quality promotions, such as coupons and deals, from your website" and apply them to campaigns. The email says this currently applies only to Search and Performance Max campaigns with linked location assets and no promotion assets, and that it is on by default with an opt-out.
Google's help page says it applies to active Search and Performance Max campaigns with location assets attached (including Local Ad campaigns) and no manual promotions; only active, unexpired offers are shown; each automated promotion runs for less than 24 hours and can be re-enabled if still verifiable on the site; promotions show "Google AI" in the Added by column; and the opt-out path is Assets → More → Account level automated assets → Advanced settings → Automated promotions Off. The help page labels "On" as recommended.
What this means
Impact read
Google writes offers it finds on your site into ads without your per-item review. The scope is limited to accounts with location assets for now, but it signals a direction: platforms generating marketing messages from page content automatically.
What it means for an independent brand
Brands with physical stores or location assets are directly affected. The risks: offers valid only for some customers (new-customer, member, regional) shown as general offers; daily-changing sale offers out of sync with ads because each automated promotion runs less than 24 hours. Online-only stores without location assets are outside the current scope.
Decision: certain; medium impact for accounts with location assets. Choose on or off before the sale rather than leaving the default.
Not yet
Avoid this week
- Do not stay on the default without checking your offer terms.
- If you want promotions in ads, do not rely on extraction; use manual promotion assets to control dates and conditions.
Decision boundary
Applies to brands using Google Ads location assets. Impact: medium. Confidence: high. The judgment changes if Google extends the scope to accounts without location assets, or starts extracting start and end dates (the help page says this is expected to change). It proves Google adds offers automatically; it does not prove automated promotions improve results.
Micro action: filter the asset report to Promotion and look for "Google AI" in Added by. If you want offers shown during the sale, upload manual promotion assets with dates. Accounts without location assets only need to confirm they are unaffected.
4. Amazon peak fees start October 15, and the holiday return deadline becomes an annual rule
According to eCommerce North America on July 14 and other trade reports of Amazon's July seller notice, 2026 holiday peak fulfillment fees apply from October 15, 2026 to January 14, 2027, averaging about $0.32 per unit (the same as 2025) across FBA, Remote Fulfillment with FBA, Multi-Channel Fulfillment (MCF), and Buy with Prime, with a separate 3.5% fuel and logistics surcharge that has no end date. We could not open Amazon's Seller Central original.
Amazon's October 16 Seller News post says items bought on Amazon.com from November 1 to December 31, 2026 can mostly be returned until January 31, 2027 (January 15 for Apple-brand products), for both FBA and merchant-fulfilled orders, with no seller action required, and that "this policy will apply every year". The post sits behind the seller login; we rely on a trade report of its text.
What this means
Impact read
Both matter to independent stores. Brands using Amazon MCF to fulfill their own-site orders pay more per order in the peak. Making the extended return window an annual rule fixes "buy for the holidays, return in January" as a default expectation for US shoppers.
What it means for an independent brand
Shopify and other independent stores fulfilling via MCF should add the peak fee and fuel surcharge to free-shipping thresholds and margin. If your own holiday return window is much shorter than January 31, shoppers comparing will notice; extending it, though, adds January return volume and cash-flow pressure.
Decision: the fee facts rest on several trade reports (medium-high confidence); the return rule rests on a trade report of the official post (medium-high). Medium impact for MCF users; a policy reference for other stores.
Not yet
Avoid this week
- Do not extend your return window to match Amazon without estimating the return cost.
- Do not estimate MCF cost with the $0.32 average alone; the actual increase varies by size, weight, and program.
Decision boundary
Applies to independent brands using Amazon MCF or selling on Amazon as well. Impact: medium. Confidence: medium to high. The judgment changes if Amazon's fee preview differs from the reported numbers, or if your category's return rate rises sharply after a longer window.
Micro action: check peak rates for lead MCF SKUs in Amazon's fee preview and multiply by expected shipments from October 15 to January 14. For your own store, write down the holiday return decision and reason (keep, extend to January 31, or extend only for gift categories) by October 28, ahead of holiday orders from November 1, and review January returns in February.
5. Shopify fills in peak fulfillment tools: batch pick and pack, automatic labels, pickup points
Shopify's merchant changelog published four fulfillment updates on October 15–16. Batch fulfillment (October 16): up to 250 fulfillments per batch, with default Pick, Print, and Pack steps. Pick prints a product list across the batch; Print buys labels for all orders at once and prints packing slips, labels, and customs forms; Pack verifies items by scanning with audio and visual feedback; steps and owners are customizable. Automatic labels in Flow (October 15): Flow buys labels for new orders matching conditions such as tags, products, or single-item orders, with carrier, service, package, weight, and ship date set in advance. Royal Mail pickup points (October 15): UK buyers can choose a Royal Mail pickup point at checkout from a map or list. Labels in the Netherlands and Poland (October 16): merchants shipping from these countries can buy DHL eCommerce labels in Shopify.
The changelogs do not state plan limits; Flow labels and batch fulfillment also do not list supported regions.
What this means
Impact read
None of these are new ideas, but they now connect inside Shopify's own admin: automatic labels, batch picking, and scan-to-pack error checks. For small and mid-sized brands that ship themselves, the peak bottleneck is often not traffic but how many orders can go out correctly each day.
What it means for an independent brand
Self-fulfilling stores shipping tens to hundreds of orders a day benefit most. Scan-to-pack can reduce mis-ships, and automatic labels cut manual work. UK stores can offer pickup points to reduce failed deliveries, and European merchants shipping from the Netherlands or Poland gain a native label option.
Decision: certain; medium impact for self-fulfilling stores. Worth trying before the peak, but run one batch on a quiet day first.
Not yet
Avoid this week
- Do not switch on batch fulfillment or automatic label rules for the first time on peak sale days.
- Do not let Flow label conditions cover orders that need a human check (high value, unusual addresses, pre-orders).
Decision boundary
Applies to small and mid-sized brands shipping from the Shopify admin. Impact: medium. Confidence: high. The judgment changes if your existing 3PL or WMS is more reliable than the native tools, or if batch label purchase is unavailable for your carrier. It proves the process can be faster; it does not prove mis-ships fall, which needs your own data.
Micro action: on a quiet day, run one batch of 20–50 orders and record total time, errors caught by scanning, and label failures; compare with manual handling time for similar orders the previous week. If results disappoint, keep the current process and revisit after the sale.
Watchlist: TikTok Shop US seller terms and Universal Cart's coverage
TikTok Shop US Seller Terms of Service. The terms on TikTok Shop's US Seller University show an effective date of October 15, 2026, covering US sellers and Japan-based cross-border sellers selling to US buyers. The page does not list differences from the previous version and points to a separate update notice, which we could not find. Existing provisions worth noting: TikTok may amend referral fees with notice by email or Seller Center; sellers may use platform data only to process and fulfill orders and not for direct marketing without consent; sellers may not solicit reviews or ask for negative reviews to be removed.
Universal Cart's coverage. Google's May I/O preview mentioned YouTube and Gmail, but the October 15 launch covers only US Search and the Gemini app. The launch post also gives no international timing, merchant fees, ranking rules, or cart data-sharing terms. Until those are published, the judgment in topic 2 stands; if Google publishes how transferred orders are attributed, or extends coverage to new surfaces or countries, we will reassess.
Impact read
What to do: brands selling on TikTok Shop US should read the update notice in Seller Center, focusing on fee and data-use terms. Independent stores selling in the US should check Google's shopping announcements and Merchant Center reports again on October 26 for wider Universal Cart coverage or separate click and order data.
The five changes on one operating chain: the execution layer arrived, the data layer has not
| Layer | This week's change | What it can answer | What it cannot prove |
|---|---|---|---|
| Discovery and cart | Shop MCP, Universal Cart | Whether agents and Google can find and cart your products | That they bring incremental orders |
| Checkout and consent | Shop Pay agent checkout, UCP buyer consent | How buyer confirmation and marketing consent work in agent orders | That agent orders can be identified and attributed |
| Offers and ads | Google Ads automated promotions | Which site offers Google writes into ads | That automated promotions improve results |
| Cost and policy | Amazon peak fees, holiday return rule | MCF cost change and US shoppers' return expectations | That longer return windows raise conversion |
| Fulfillment | Shopify batch fulfillment, auto labels, pickup points | Whether orders ship faster and more accurately each day | That mis-ships certainly fall |
Impact read
The first three rows say the same thing: agents and platforms can now read and act on your products, prices, offers, and consent settings. The execution layer arrived faster than the data layer; there are no official reports yet on identifying agent orders, attributing transferred orders, or measuring automated promotions.
So the sensible pre-sale move is "make settings correct and keep definitions reviewable": accurate prices, stock, and codes; consent defaults that fit each market; automated features explicitly on or off. Decide on expansion after the sale based on data you can read back. Counter-signals: an AI assistant returns prices or stock that differ from the store; promo codes appear where they should not; automatic label rules buy labels for orders that needed a human check. If any appear, fix settings before expanding channels.
Baseline execution checklist: five optional actions, finish one first
Action card
Ask an AI assistant about your lead products
Use an MCP-capable AI assistant to find five lead products and record whether title, price, discount, stock, and return information match the store.
- Applies to
- Merchants with products in Shopify Catalog
- Done when
- All five products have a comparison record, and mismatches have an owner.
- Owner
- Product operations
- Effort
- 30–45 minutes
- Stop when
- Stop if the test would create a real order or charge.
- Review at
- 10/23/2026, 2:00:00 PM
Related resources
Action card
Inventory publicly usable promo codes
List every promo code that works publicly; add conditions to, or disable, creator-specific, internal test, and expired-campaign codes.
- Applies to
- Independent stores selling in the US with promo codes
- Done when
- Each active code has a purpose, audience, and expiry, and codes that should not be public are handled.
- Owner
- Marketing lead
- Effort
- 45 minutes
- Stop when
- If disabling a code affects an ongoing partnership, confirm with the partner first.
- Review at
- 10/23/2026, 2:00:00 PM
Related resources
Action card
Decide on Google Ads automated promotions
Check promotions added by "Google AI" in the asset report; if you want offers shown in the sale, use dated manual promotion assets; otherwise switch it off in account-level automated assets.
- Applies to
- Google Ads accounts with location assets
- Done when
- The account has an explicit on or off decision, and sale promotion assets are uploaded manually.
- Not for
- Accounts without location assets
- Owner
- Paid media lead
- Effort
- 30 minutes
- Stop when
- Switch off immediately if an automated promotion conflicts with on-site offer terms.
- Review at
- 10/21/2026, 2:00:00 PM
Related resources
Action card
Set the holiday return window
Compare your current return policy with Amazon's January 31 rule, estimate the return volume and cash-flow effect of a longer window, and write down the decision to keep, extend, or extend for gift categories only.
- Applies to
- Independent stores selling to US buyers
- Done when
- A written decision and reason exist and are reflected in return settings and support scripts.
- Owner
- Ecommerce lead and finance
- Effort
- 60 minutes
- Stop when
- If the return cost cannot be estimated, keep the current policy.
- Review at
- 10/28/2026, 2:00:00 PM
Related resources
Action card
Try one batch fulfillment run on a quiet day
On a non-peak day, process 20–50 orders with Shopify batch fulfillment and record time, errors caught by scanning, and label failures; compare with last week's manual handling.
- Applies to
- Stores shipping from the Shopify admin
- Done when
- A time and error comparison exists, with a decision on using it during the sale.
- Not for
- Stores fully outsourced to a 3PL
- Owner
- Warehouse or shipping lead
- Effort
- Half a day
- Stop when
- Return to the current process if batch label purchase fails or scanning slows shipping.
- Review at
- 10/30/2026, 2:00:00 PM
Related resources
Priorities at the end: lock offers and policies, then look at agent channels
| Priority | Who should look first | Action | Why |
|---|---|---|---|
| P1 | Lock | Google Ads accounts with location assets; US stores using promo codes | Decide on automated promotions; inventory public promo codes | Platforms are already surfacing your offers automatically |
| P1 | Decide | Stores selling to US buyers; stores fulfilling via MCF | Set the holiday return window; calculate MCF peak cost | Return expectations and fulfillment cost have changed |
| P2 | Check | Merchants in Shopify Catalog; self-fulfilling stores | Check lead products via an AI assistant; trial batch fulfillment on a quiet day | Data accuracy and shipping capacity cap the sale |
Avoid this week
- Do not enter the sale with Google automated promotions left on default.
- Do not rewrite product information in bulk before the sale for agent channels.
- Do not extend returns to January 31 without estimating the cost.
- Do not switch on a new fulfillment process for the first time on a peak day.
Sign in to continue with the analysis and execution cards available to your plan
You are viewing the Public edition. The server returns the sections available to your account, while unopened content stays out of public APIs and page source.
View plan access