Ecommerce Weekly | AI channels now ask merchants for facts they can sign off
26 October – 1 November 2026: Shopify Fall '26 knowledge base and channel mapping, UCP reports, AI Mode ads, Meta product browsing, app token deadline and local payments
This week's densest changes came from Shopify's Fall '26 Edition and Google's ads and Merchant API updates: the store facts agents use to answer shoppers, the product data sent to each channel and the reports after an agent-channel sale all got official entry points. At the same time AI Mode ads reached ten countries, Meta switched on product browsing in existing ads by default, and Shopify set a hard deadline for public-app token migration. Each item separates confirmed facts, what they prove, what they do not yet prove and what an independent store can do first. Guest and Basic receive the same facts and baseline actions; Pro and Max add trade-offs by brand type, counter-signals and a seven-day board.
Editorial linePublicThe week in one line: AI channels now ask merchants for facts they can sign off
Open the editorial line and its boundaries
The densest changes of 26 October to 1 November came from Shopify's Fall '26 Edition (published 28 October) and Google's ads and Merchant API updates the same week. Read together, platforms are filling in three things: the store facts agents use to answer shoppers, the product data sent to each channel, and the reports that follow an agent-channel sale. In the same week Google expanded AI Mode ads to ten countries, Meta switched on in-app product browsing for existing ads by default, and Shopify set a hard deadline for public apps to move to expiring tokens.
What these updates share is that platforms hand "can an agent understand and buy this?" back to the merchant for sign-off, and hand over the review and readback work with it. Facts a merchant approves are used directly by agents to answer customers; gaps in product data become display gaps on a specific channel.
Verified facts
- On 26 October Shopify built the knowledge base into the admin (Sales channels > Agentic). Shopify drafts store facts from storefront pages, and nothing reaches AI agents until the merchant publishes it; the standalone Knowledge Base app stops after 15 January 2027.
- The Fall '26 Edition lists selling in Google Search (including AI Mode) and the Gemini app with Shopify checkout, a Catalog eligibility view and a UCP /ask endpoint; the page gives no region or plan labels.
- On 28 October Shopify added a page to map product data per sales channel, with previews and field coverage percentages; on 29 October the admin began suggesting GTINs checked against the GS1 registry.
- Google's October Merchant API update adds two v1alpha report views for UCP agent-channel funnels, orders, cancellations and buy-button clicks by date and customer country; Shopify storefronts gained UCP /buy links the same day.
- On 28 October Google expanded AI Mode ads to ten countries including Australia, Canada, New Zealand and Singapore, and launched multiturn ads in follow-up queries in the US.
- From 28 October Meta switched on product browsing in existing ads that had not opted out, so a tap may first open an in-app product page on Facebook or Instagram.
- On 27 October Shopify set the timeline: public apps must use expiring offline tokens before 1 January 2027 or be delisted and blocked from new installs, with affected requests returning 403 from 1 April 2027.
- Shopify Payments added seven local methods: Alipay, WeChat Pay, MB WAY, OXXO, Satispay, Scalapay and GrabPay.
Start with three things: review the draft store facts in the knowledge base, check field coverage for your main channels, and confirm the vendors of critical apps are ready for the token migration.
Impact read
Macro view: agent channels are moving from "can we connect" to "what is said, what is sold and how it is counted". The facts layer (knowledge base), the data layer (channel mapping, GTIN), the transaction layer (/buy, checkout inside Google) and the measurement layer (UCP reports) all got official entry points this week, but measurement is still alpha and the transaction layer's availability is not spelled out. The job for independent stores this week is to get facts and data right, not to carve out a separate agent-channel budget.
1. Shopify's knowledge base moves into the admin: merchants sign off on what agents say
On 26 October Shopify built the knowledge base into the admin under Sales channels > Agentic > Knowledge base. It is one page to review, edit and publish store facts such as contact details, gift services, return windows and shipping destinations. Shopify drafts these from published storefront pages, each draft linking to its source, and the update states that nothing Shopify drafted reaches AI agents until the merchant publishes it. The knowledge base holds up to 1,000 facts. Former Knowledge Base app users with fewer than 1,000 facts were migrated and published; those with more must re-add facts before the app stops on 15 January 2027.
The Fall '26 Edition page of 28 October lists it alongside two related items: selling in Google Search (including AI Mode) and the Gemini app with Shopify-powered checkout, and a UCP /ask endpoint that answers questions about a store grounded in facts the business publishes. The page gives no region or plan labels for either.
What this means
Impact read
When an agent answers "how long will it take?" or "can I return it?", it can now draw on the version the merchant signed off in the admin rather than only what it scraped. The upside is control; the risk is that an outdated or inconsistent approved fact gets repeated by agents with confidence.
What it changes for independent stores
Cross-border stores with complex or market-specific returns, shipping, duties or warranty policies are most affected. Former Knowledge Base app users should confirm migrated facts are still accurate. Stores that never used it have nothing published automatically, which also means agent channels have no reviewed layer for them yet.
Judgment: a high-relevance, high-certainty change. It does not create orders directly, but it decides how the store is described in agent channels.
What to do this week
Decision boundary
The support lead spends 45 minutes on the drafts and publishes only facts that match the wording of a policy page and are clear per market (return window, shipping countries, delivery estimates, whether duties are included); everything else stays unpublished. Done when every published fact maps to a storefront page and the support team confirms it matches day-to-day replies. Review whenever return or shipping policy changes.
Not yet
Avoid this week
- Do not publish all Shopify-drafted facts in one click.
- Do not publish temporary Black Friday policies without an end date.
- Do not assume selling inside Google is open to every store; go by what the admin shows.
2. Per-channel product data mapping and GTIN checks: you can now see what each channel receives
On 28 October Shopify launched a channel mapping page: preview the product data sent to each sales channel; fields such as title, images, category and colour have default sources that can be repointed or combined (for example adding the brand name to the title); override any field for one product on one channel; bulk edit; and see a coverage percentage for each field. The update does not list specific channels or eligibility.
On 29 October the admin began checking barcodes against the GS1 registry: barcodes that look like registered GTINs are labelled GTIN; saved Custom barcodes that look like GTINs get a suggestion; hovering shows the GS1 record for comparison with vendor and product name; and suggestions can be accepted or dismissed in bulk. All of this is informational, and saved barcodes do not change unless a suggestion is accepted.
What this means
Impact read
Sending one identical product record to every channel is no longer enough. Channels, including agent channels, read different fields with different rules, and the gaps are now visible in one place. GTIN checks turn an often-ignored field into something reviewable; it is the same barcode field used for the EU customs identifiers covered in W43.
What it changes for independent stores
Stores selling across Google, Meta, TikTok, marketplaces and agent channels benefit most. Own-brand products without registered GTINs are not forced to change, but barcode types should be labelled correctly so internal codes are not marked as GTINs.
Judgment: medium-to-high impact, high certainty. It does not improve ranking by itself, but it sharply shortens the search for why a channel is not showing a product.
What to do this week
Decision boundary
The product-data owner opens channel mapping, picks the highest-traffic channel and records coverage for title, category, colour and barcode, then checks the preview of 20 best sellers against their product pages. In the bulk editor, review GTIN suggestions for those 20 and accept only those consistent with the GS1 record, the supplier and the product name. Done when there is a channel × field coverage table and a processed GTIN list.
Not yet
Avoid this week
- Do not change title mappings for every channel right before the sale.
- Do not accept all GTIN suggestions at once; check the GS1 record and the actual supplier first.
3. Agent channels get reports: new UCP views in the Merchant API, /buy links on Shopify
Google's October Merchant API update (page last updated 28 October) adds two v1alpha views to the Reports sub-API: UcpAgenticPerformanceView with account-level conversion funnel, order, cancellation and cart-transfer metrics segmented by date and customer country, and UcpAgenticProductPerformanceView with product-level (offerId) buy-button clicks segmented the same way, across Google surfaces such as AI Mode and Gemini. The same update adds commission fields to YouTube Shopping affiliate reporting, a videoLinks product attribute and warranty attributes for all products.
On 28 October Shopify announced storefront support for the UCP permalink capability on a new /buy route: one link can add items, prefill buyer details and open checkout, and stores advertise the endpoint at /.well-known/ucp. Existing /cart links keep working; /buy does not yet support selling plans, line item properties, cart notes or cart attributes.
What this means
Impact read
For the first time, agent channels have official data split by product and country. It is alpha, so fields and definitions may change, and it covers only Google surfaces that run through UCP. /buy gives external agents and link tools a standard route straight to checkout; ordinary merchants do not need to do anything.
What it changes for independent stores
Merchants already on Google UCP (cart transfer or native checkout), or teams with developers on the Merchant API, can start a baseline. Others need no action this week, beyond knowing that agent-channel value will be judged from funnels like these, split by country, rather than from impressions or citations.
Judgment: medium impact, medium certainty (alpha). Build a read-only baseline; do not set budgets from it.
What to do this week
Decision boundary
For teams on UCP, the data lead pulls both views for the last 14 days through the Merchant API, records cart transfers, orders and cancellations by customer country, and reconciles once with Shopify orders for the same period. Done when a baseline table records field definitions, time zone and gaps. Teams not on UCP skip this week.
Not yet
Avoid this week
- Do not allocate a separate agent-channel budget from alpha data.
- Do not rewrite every /cart link for /buy now; keep /cart for links that use selling plans or line item properties.
4. Google AI Mode ads reach ten countries and follow-up queries in the US: cross-border accounts should watch search terms
On 28 October Google announced on its blog that ads in AI Mode are scaling in Australia, Canada, India, Kenya, Malaysia, New Zealand, Nigeria, Pakistan, the Philippines and Singapore, with more locales including Brazil, Mexico and the UK planned for Q4 2026, and that multiturn ads launch in the US so ads can appear in follow-up queries in AI Mode, labelled Sponsored. Citing its own data, Google says AI Mode queries are on average three times longer than traditional searches and that sponsored and organic product list items in AI Mode already have similar click-through rates. Trade coverage adds that existing Search, Shopping and PMax campaigns are automatically eligible with no separate budget, that AI Mode cannot be targeted or excluded separately, and that these queries appear in aggregate in standard and search-terms reports.
What this means
Impact read
For advertisers this is not a new channel to switch on but a change in the traffic mix of existing campaigns: longer, more specific, context-rich queries enter the same auction. Because reporting is aggregated, the effect of AI Mode is hard to isolate.
What it changes for independent stores
Cross-border brands running Shopping or PMax in those ten countries, especially Australia, Canada, New Zealand and Singapore, are most affected. How well titles, attributes and landing pages match long queries matters more than before. The UK, Brazil and Mexico are planned for Q4 with no dates.
Judgment: a medium-impact, high-certainty distribution change; certainty about its effect is low because it cannot be reported separately.
What to do this week
Decision boundary
The paid-media lead compares search-terms reports for core campaigns in the affected countries for seven days before and after 28 October: share of queries of six or more words, CTR and conversion rate, and adds costly irrelevant long queries as negatives. Done when there is a before/after table and a negative-keyword log. Review on 12 November.
Not yet
Avoid this week
- Do not raise budgets because of Google's statement that AI Mode CTRs are similar to organic.
- Do not create separate campaigns for AI Mode; it cannot be targeted on its own.
5. Meta turns on product browsing in existing ads: realign site-traffic and landing-page test definitions
According to Meta's in-product notice as quoted by PPC Land on 15 October, Meta switched on product browsing in existing ads from 28 October unless advertisers opted out before then. With it on, a tap may first open a page inside Facebook or Instagram with product information and personalised suggestions before the website. After the deadline it can still be turned off per ad in Advantage+ creative settings or through bulk edit. Meta's notice says an experiment from 6 to 21 August across about 117,700 advertisers showed 3% more conversions on average; the figure is Meta's own and the metric is not defined. At verification no matching Meta Help Center or newsroom page was found.
What this means
Impact read
An in-platform page now sits between the click and the visit. Some people will browse inside Meta and leave, or enter the site on a different product. The gap between ad clicks and site sessions can widen, and landing-page A/B tests get diluted by this extra layer.
What it changes for independent stores
Brands running Sales objectives that depend on specific landing pages (pre-order, bundle or review pages) are most affected. Catalogue-led accounts that already let Meta choose products are less affected.
Judgment: medium impact; the default switch-on is medium-to-high certainty (several trade outlets quote the same notice), the performance figure is low certainty.
What to do this week
Decision boundary
The paid-media lead checks Advantage+ creative settings on ten live ads, records whether product browsing is on, and turns it off for ads that rely on a dedicated landing page. Then compare link clicks / landing-page views / site sessions for seven days before and after 28 October. Done when there is a settings list and a before/after ratio table. Review on 11 November.
Not yet
Avoid this week
- Do not adopt the "3% more conversions" figure as your own expectation.
- Do not change landing pages and this setting at the same time during the sale, or you will not know which one mattered.
6. A hard deadline for Shopify public-app tokens: migrate before 1 January 2027 or be delisted
On 27 October Shopify marked an "action required" developer changelog entry: existing public apps must move to expiring offline access tokens before 1 January 2027. The timeline: from 1 January non-compliant apps are removed from the App Store and blocked from new installs, while existing installs can keep calling until 1 April; from 15 January merchants using non-compliant apps see admin warnings about possible interruption; from 1 April affected Admin API requests using non-expiring offline tokens return HTTP 403. Custom apps and merchant-created apps are not affected. The same day Shopify introduced an optional IP allowlist so developers can restrict Admin API tokens to their own server addresses.
What this means
Impact read
This is the first concrete timeline Shopify has published after the app-security incidents covered in W43. With short-lived tokens, a leaked token can be abused for a shorter window. For merchants the risk is that a vendor they depend on does not keep up and is delisted early next year, or breaks after April.
What it changes for independent stores
Stores relying on several public apps for orders, inventory, shipping, reviews or marketing are most affected, especially small vendors or apps that have not been updated in a long time. Custom apps you built yourself are out of scope.
Judgment: medium impact, high certainty for app-dependent stores. With two months to go, this is time for a survey, not replacements.
What to do this week
Decision boundary
The store owner lists the critical apps for orders, inventory, shipping and payments, checks each vendor's changelog or asks directly whether they have moved to expiring tokens, and records the expected date. Done when there is a migration-status table for critical apps. If a vendor has no clear answer by 15 December, start evaluating alternatives.
Not yet
Avoid this week
- Do not replace critical apps during Black Friday because of this deadline.
- Do not read "the app still works" as compliance; delisting and 403 errors arrive in stages.
7. Shopify Payments adds seven local methods: one fewer payment gap in cross-border checkout
The Fall '26 Edition of 28 October announced seven new local payment methods in Shopify Payments: Alipay, WeChat Pay, MB WAY, OXXO, Satispay, Scalapay and GrabPay. All work in online checkout, and most also work with Shop Pay, the Shop app and Shopify POS. The Editions page lists Alipay, WeChat Pay, Satispay, Scalapay and GrabPay for in-person use, with MB WAY and OXXO online. It does not state, method by method, which merchant countries are eligible, and trade write-ups note that some methods are still rolling out in some countries. The same Edition mentions multi-currency payouts in Australia, France, Singapore and Hong Kong, limited to Advanced and Plus plans.
What this means
Impact read
Most of these methods used to need a third-party payment plug-in. Inside Shopify Payments, activation, reconciliation and disputes stay in one admin. Each maps to a market habit: MB WAY for Portugal, OXXO for Mexico, Satispay and Scalapay for Italy and nearby, GrabPay for Southeast Asia, and Alipay and WeChat Pay for Chinese travellers and customers.
What it changes for independent stores
Stores with meaningful orders from Portugal, Mexico, Italy or Southeast Asia that currently accept only cards are most affected. Stores without those markets do not need to change checkout.
Judgment: medium impact where the markets match; high certainty about the feature, medium certainty about country availability.
What to do this week
Decision boundary
The cross-border lead reviews 90 days of checkout abandonment and order share by country, picks one market that matches these methods and checks under Settings > Payments whether it can be enabled. If so, enable one and watch checkout completion, refunds and disputes in that market for two to four weeks. Done when there is a record reading enabled, not available in our country, or not needed.
Not yet
Avoid this week
- Do not launch a new payment method for the first time during sale week.
- Do not enable several new methods at once, or you will not know which one mattered.
Watchlist: Meta's terms, Merchant Center passkeys and the first EU invoices
Verified facts
- Meta's terms (follow-up to W43): the UK CMA said on 30 October that Meta had announced on 28 October a pause to the rollout of its new commercial terms pending consultation with industry groups, and opened a Competition Act 1998 investigation the same day, with no finding of wrongdoing. Yet at verification on 1 November Meta's Self-Serve Ad Terms page still said the update takes effect on 20 November. The two conflict; no conclusion yet. Next check: 18 November.
- Google Merchant Center passkeys: the help page says sensitive actions such as changing the website URL or managing users require passkey confirmation, rolling out gradually to all users; "as soon as 29 October" comes from Google's email to account owners. Agency or multi-user accounts should confirm the owner's device has a passkey. Next check: 8 November.
- EU handling fee (follow-up to W43): applies from 1 November. No new official change this week; still waiting for the first carrier invoices to confirm the charging unit. Next check: 9 November.
The seven changes on the agent and ads chain: which layer got an entry point, which is still empty
| Layer | This week's change | What it answers | What it does not prove |
|---|---|---|---|
| Store facts | Shopify knowledge base, UCP /ask | Which version of policy agents use to answer shoppers | That agents will cite those facts |
| Product data | Channel mapping, GTIN suggestions | What each channel receives and which fields are missing | That filling fields brings more impressions |
| Transaction | Shopify checkout inside Google, /buy links | Where agents can send shoppers to pay | That every store has access |
| Measurement | Merchant API UCP views (alpha) | Agent-channel funnel, orders and cancellations | That those orders are incremental |
| Paid distribution | AI Mode ads in ten countries, Meta product browsing on by default | Where traffic enters and which layer it passes through | Conversion and margin of the new traffic |
| Platform security and payments | App token deadline, seven local methods | Which dependencies break when, which markets lose a payment gap | That a payment method lifts conversion |
Impact read
The table shows that every layer of the agent channel now has an entry point, while the layer that would prove value, measurement, is still the thinnest. What is possible and worth doing now is getting the first two layers right: published facts that match policy pages, and field coverage for the main channels that you can state. Neither depends on proof that a new channel works, so neither is wasted effort.
The two paid-distribution changes (AI Mode ads, Meta product browsing) share one pattern: platforms add more of their own surfaces between the click and the visit, while reports are not split accordingly. Judge them by site-side sessions, add-to-carts and contribution margin, not platform-side CTR.
Decision boundary
Counter-signals: if support tickets about "the agent said something wrong" rise after publishing knowledge-base facts, withdraw ambiguous facts first. If long-query share rises after the AI Mode expansion while conversion falls, add negatives before changing bids. If site sessions fall after Meta product browsing turns on while orders hold, part of the browsing is being absorbed in-app; read it with your attribution definition rather than sessions alone.
Baseline checklist: five optional actions, finish the first one first
Action card
Review knowledge-base drafts and publish only facts that match policy pages
Go through Shopify-drafted facts under Sales channels > Agentic > Knowledge base and publish only return, shipping and contact facts that match a policy page and are clear per market.
- Applies to
- All Shopify stores, especially cross-border and multi-market stores
- Done when
- Every published fact maps to a storefront page and support confirms it matches daily replies.
- Owner
- Support lead
- Effort
- 45 minutes
- Stop when
- If policy pages contradict each other, fix them first and hold the related facts.
- Review at
- 11/9/2026, 3:00:00 PM
Related resources
Action card
Record field coverage for one main channel
Open channel mapping, pick the highest-traffic channel, record coverage for key fields, and check previews and GTIN suggestions for 20 best sellers.
- Applies to
- Multi-channel Shopify stores
- Done when
- A channel × field coverage table exists and GTIN suggestions for 20 products are reviewed.
- Owner
- Product-data owner
- Effort
- 60 minutes
- Stop when
- If mapping changes would affect ads running in the sale, wait until after it.
- Review at
- 11/12/2026, 3:00:00 PM
Related resources
Action card
Check product browsing on Meta ads
Check Advantage+ creative settings on ten live Sales ads, turn product browsing off for ads that rely on a dedicated landing page, and record click-to-session ratios for seven days before and after.
- Applies to
- Brands running Sales-objective ads on Meta
- Done when
- A settings list and a before/after ratio table exist.
- Owner
- Paid-media lead
- Effort
- 30–45 minutes
- Stop when
- If the window overlaps the sale and causes cannot be separated, record without concluding.
- Review at
- 11/11/2026, 3:00:00 PM
Action card
Compare search terms in the AI Mode expansion countries
For core campaigns in affected countries such as Australia, Canada, New Zealand and Singapore, compare search-terms reports for seven days before and after 28 October and handle costly irrelevant long queries.
- Applies to
- Cross-border brands running Shopping or PMax in those countries
- Done when
- A before/after table of long-query share, CTR and conversion rate plus a negative-keyword log.
- Not for
- Accounts not serving those countries
- Owner
- Paid-media lead
- Effort
- 45 minutes
- Review at
- 11/12/2026, 3:00:00 PM
Action card
Build a token-migration status table for critical apps
List public apps that handle orders, inventory, shipping and payments; check vendor updates or ask whether they have moved to expiring offline tokens.
- Applies to
- Shopify stores relying on third-party public apps
- Done when
- Each critical app has a migration status and expected date.
- Not for
- Stores using only custom apps
- Owner
- Store owner
- Effort
- 30 minutes
- Stop when
- Apps with no answer by 15 December move to an alternatives review.
- Review at
- 12/15/2026, 3:00:00 PM
Related resources
Dates to keep
Public apps must use expiring tokens
Non-compliant public apps are removed from the App Store and blocked from new installs; existing installs can keep calling until 1 April 2027.
Knowledge Base app ends; admin warnings for non-compliant apps
The standalone Knowledge Base app stops, so merchants with more than 1,000 facts must re-add them before then; merchants using non-compliant apps start seeing admin warnings.
Requests with non-expiring offline tokens return 403
Affected Admin API requests using non-expiring offline tokens start being rejected.
Priorities at the end: get facts and data right first
| Priority | Who looks first | This week's action | Why |
|---|---|---|---|
| P1 | Review | Support and operations at every Shopify store | Review knowledge-base drafts; publish only facts that match policy pages | Published facts are used directly by agents to answer shoppers |
| P1 | Check | Brands advertising on Meta and Google | Check Meta product browsing; compare search terms in AI Mode countries | Both add platform surfaces between click and visit without split reporting |
| P2 | Survey | Multi-channel brands and app-dependent stores | Record main-channel field coverage; build the critical-app token table | Data gaps become display gaps; app deadlines arrive in stages |
Avoid this week
- Do not publish every Shopify-drafted store fact in one click.
- Do not set budgets from alpha reports or platform-reported effect sizes.
- Do not change landing pages, payment methods and ad settings at the same time during the sale.
- Do not replace critical apps during Black Friday because of the token deadline.
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