The briefing of 30 September 2026

Shopify turned on a Google direct checkout by default for eligible stores, and says Google Analytics and custom pixels will not fire inside it.

Opened here. Nothing proves which version was read. How this is sourced

Before this. Somebody who found your product in an AI chat had to come to your own site to buy it. The sale happened on your checkout, which meant your analytics, your pixels, your upsells and your rules all ran. Shopify calls these AI chat channels agentic storefronts, and until now the Google one sent the shopper to you.

• Shopify says the default is on, and says it twice in its own words, once at the top of the page and once in the instructions for turning it off. Purchasing in direct checkouts is activated by default for eligible stores, but you can deactivate direct checkout and have customers complete their purchase on your online store instead.
• It has to stay on for anything to be bought. Shopify does not present the default as a starting position a merchant then tunes. When your store is eligible, purchasing through Google AI Mode and Gemini's direct checkout is active by default in your Shopify admin, and it has to remain active for your products to be directly bought by customers in Google AI Mode and Gemini.
• The default is not the same for every plan, and the plan Shopify names is not the one the proposal named. Shopify draws the line between eligible stores, where it is on, and stores on its Agentic plan, where it is off. Agentic storefronts setup is different with the Agentic plan, and direct checkout is deactivated by default.
• Client-side measurement does not run in that checkout, and Shopify's own sentence is broader than the two things it names first. Two server-to-server events are what fires. Google Analytics and custom pixels won't fire in Google AI Mode and Gemini's direct checkout. The checkout fires only server-to-server pixels (started, completed), so none of the standard or custom client side pixels fire.
• Both halves are still arriving, which the page says at the top before it says anything else. Google AI Mode and Gemini is rolling out to eligible stores through Shopify's agentic storefronts. Direct checkout is also rolling out to Google users and might not yet be available in your store.
• Eligible is a list, not a description. Shopify requires a store based in the United States that sells to United States customers, a valid Google Merchant Center account meeting all of Google's own requirements, products present in Merchant Center and eligible for Shopify Catalog, agreement to a supplemental terms of service, and a completed terms of service, privacy policy and returns policy. The checkout is shown only to customers in the United States.
• A good deal does not work inside it besides the pixels. Subscriptions, product bundles, customisable products and business-to-business-only products cannot be bought. Local delivery, pickup in store and pickup points are not offered. You cannot require a shopper to sign in. Only some of your accelerated checkout options appear, and payment is by credit or debit card.
• Turning it off takes two steps. Shopify's own instructions require a merchant to deactivate Allow Shopify to manage for me before the direct checkout setting can be changed at all.
• Google describes the same checkout more narrowly than Shopify does. Its own Merchant Center help says the feature is for select merchants, describes an early access programme reached through an interest form, and says only listings carrying a particular product attribute show a Buy button. The checkout feature enabled by UCP is available for select merchants at this time.
• This is not a Google-only arrangement. Shopify sells the same thing to four AI chat channels under one programme, and the direct checkout described above is what three of them use. Shopify's agentic storefronts lets customers discover and purchase your products in AI channels, such as ChatGPT, Google AI Mode and Gemini, Microsoft Copilot, and Meta.

What it means. If you sell on Shopify and you meet the eligibility list, the setting that lets somebody buy from you inside Google AI Mode or Gemini is already on, and you did not switch it on. The thing to check this week is not whether to allow it. It is what your reporting now misses. Shopify says plainly that Google Analytics and custom pixels do not fire in that checkout, and that only two server-to-server events do, started and completed. So a sale can happen, be real, be yours, and be invisible to the measurement your budget decisions come out of. Turning it off is also two steps rather than one, because you have to stop Shopify managing the channel for you first.

• your analytics team If a sale completes inside Google AI Mode, which of our reports sees it, and what do we lose by having a started and a completed event and nothing in between?
• your ecommerce team Is Allow Shopify to manage for me switched on in our admin, and did anybody in this business decide that or did it arrive?

Happened 22 September 2026Published 29 September 20267 days later

Touches 3 modules across 1 book.

IAB Tech Lab published a draft standard for the stage between a buyer's brief and a buy commitment, open for comment until 22 October.

Opened here. Nothing proves which version was read. How this is sourced

Before this. When an agency wants to buy space from a publisher, it sends a brief and gets a proposal back. That part is still done by people, in emails and spreadsheets and phone calls. Standards already cover the buying and the delivery either side of it. Nothing covered the middle.

• AAMP 3.0 and OpenProposal are not two competing things and not one thing. The body calls AAMP 3.0 the next major release of its Agentic Advertising Management Protocols framework, and OpenProposal the new specification inside that release. AAMP 3.0 addresses that gap through OpenProposal, a new specification that defines how sellers can represent advertising products and respond to buyer briefs in a standardized, agent-readable format.
• The comment window closes on 22 October 2026. The date is on the press release four times, in its subheading, in its own page description, in the body of the text and in the closing line that tells a reader where to go. It is on the body's GitHub release as a date range and in the title of its own tracking issue. The new specification will be in Public Comment until October 22, 2026, during which time it will be built into the Buyer and Seller Agent SDK reference implementations.
• It stops at the buy commitment. The body lists seven stages, from a seller publishing a catalogue of packages through discovery, comparison, qualification and negotiation, and then hands the transaction to the standards already in use. It says existing programmatic implementations can carry on unchanged. Existing programmatic implementations of those standards can continue operating as they do today, without requiring DSPs, SSPs or other technologies to change their existing use cases.
• The working group says seven things are undecided and names them in public, including that the format registry has no entries for audio, connected TV or outdoor, and that comparing guarantees across channels is unsolved. The working group is undecided on these. They are the most useful place to comment.
• One national body has opened a seller side pilot. Its stated position is a sandbox before any live money moves. Our role now is to help Australian publishers test discovery, proposal and approval workflows in a sandbox before any live spend is involved.
• The release also hardens the agent software kits rather than only adding the specification. It names guardrails on pricing, packages, orders and buy details, trust in agent identity, and making a retry safe. The body explains the last one in plain words itself. In practical terms, idempotency means an instruction can be sent twice without happening twice; so when an agent retries a request it didn't get an answer to, the result is confirmation of the original order rather than a duplicate.

What it means. There is now a draft you can name in a conversation with a seller, and there is not yet anything you can hold one to. If a platform tells you this quarter that its planning agent is standards based, the useful question is which draft, at which tag, and what it does when the answer is no. The comment window closes on 22 October, which is the last month in which a buyer can put a requirement into the text rather than discover it afterwards.

• your planning team If a seller sends us a proposal an agent can read, who here reads the terms behind the price, and would we notice if the reasoning attached to it was wrong?
• your programmatic team Do we have anybody who could answer the seven open questions from our own experience, and are we going to file that before 22 October or after it is decided?

Happened 22 September 2026Published 29 September 20267 days later

Touches 3 modules across 1 book.

The Supreme Court stayed the ruling that had removed the lowest broadcast rates from party committees, rather than deciding whether it was right.

Opened here. Nothing proves which version was read. How this is sourced

Before this. A broadcaster has to sell a political candidate advertising at its lowest rate in the weeks before an election. A regulator said in March that party committees working alongside candidates were entitled to that rate too. An appeal court struck the regulator down in August, and the committees asked the Supreme Court to put that ruling on hold while they appealed.

• On 4 September 2026 the Court granted a stay in National Republican Congressional Committee v. Brown, No. 26A274, recalling and staying the Fourth Circuit's mandate pending the filing and disposition of a petition for certiorari.
• The stated ground was a strong likelihood that the Fourth Circuit lacked statutory jurisdiction, because an application for review to the Commission is a condition precedent to judicial review under 47 U.S.C. section 155(c)(7).
• The Court also found likely irreparable harm, recording the committees' representation that broadcasters were already rescinding favourable rates.
• The underlying dispute is a Media Bureau public notice of 30 March 2026 stating that candidates, joint fundraising committees with noncandidate members, and parties engaged in coordinated activity were all entitled to the lowest unit charge. A divided Fourth Circuit panel set that notice aside on 25 August 2026.
• Justice Jackson dissented and would have denied the stay, on the view that the applicants are unlikely to succeed on the jurisdictional argument.
• If certiorari is denied the stay terminates automatically. If it is granted, the stay terminates when the Court's judgment issues.

What it means. For this cycle, protected demand keeps its protected price in the window when local inventory is tightest. It is interim rather than settled: the ruling that removed the access has been stayed, not overturned, and the stay ends by its own terms if the Court declines the case. Plan for the access being there, and do not treat the question as closed.

The Big Question this moves. How do you hold local reach when protected political money takes the inventory?

• your local investment team In which of our markets does the political window put our reach at risk this cycle, and what have we committed early enough to protect?

Happened 4 September 2026Published 24 September 202620 days later

Touches 2 modules across 1 book.

The FTC and 22 states sued Amazon, alleging its advertising auction carried an undisclosed surcharge.

Opened here. Nothing proves which version was read. How this is sourced

Before this. Amazon sells the advertising space beside its own search results. For years it told advertisers that the winner of each auction pays one penny more than the next highest bid, so a bidder could name what the space was worth to them without fear of paying it. More than a million brands and sellers bid on that basis.

• The complaint was filed on 31 August 2026 by the Federal Trade Commission and 22 state attorneys general in the Western District of Washington, as case 2:26-cv-03097.
• It is brought under Section 5(a) of the FTC Act, which prohibits unfair or deceptive acts or practices, together with state consumer protection statutes, and seeks a permanent injunction and a monetary judgment.
• It alleges that Amazon told advertisers it ran a second price auction in which the winner pays one cent more than the next highest bid.
• It alleges that the share of the time Sponsored Products advertisers were charged their own bid rose from between 30 and 40 per cent in 2021, to 70 per cent in 2022, and to about 80 per cent in 2024, as a result of the surcharges. increased from between 30% and 40% in 2021, to 70% in 2022, and to approximately 80% in 2024
• It alleges that an undisclosed surcharge, referred to internally as a soft reserve price, was added to the auction from 2019 without notice.

What it means. The complaint alleges that a retail media price contained something the buyer was not told about. If any of it is proven, the question stops being whether the auction prices competition and becomes whether the auction is the price at all. None of it has been proven yet.

The Big Question this moves. Does the retail media auction price competition, or your dependence on the shelf?

• your retail media team For each network we buy, what do we actually know about how the price was set, and where are we accepting a number with no way to check it?
• your retail media team If the share of auctions where we simply paid our own bid has been climbing since 2021, would our own reporting have shown us that, and can we ask for it?

Happened 31 August 2026Published 24 September 202624 days later

Touches 2 modules across 1 book.

A standards body finalised its containerisation specification, adding a layer that lets one container call another and merge the answers into one bid.

Opened here. Nothing proves which version was read. How this is sourced

Before this. A container is a sealed box holding somebody else's software, running on machines you own. A standards body has spent a year writing down how one company can run its code inside another company's advertising auction that way. Until August what existed was a draft open for comment, and a draft is not something a buyer can hold a supplier to.

• The working group that produced this opened in June 2025 under the name containerisation, and the framework it became was released for public comment on 13 November 2025.
• The finished version was declared final on 12 August 2026.
• The addition over the comment draft is a layer that allows one container to call others, potentially run by other companies, and merge their answers into a single change to one bid request.
• The headline performance figure attached to this work, a reduction in bid request and response time of up to 80 per cent, is written by the body as what the framework is designed to do. No baseline, no method and no test result has been published for it.
• A seller now restates that same figure on a product page as a reduction achieved, crediting the standards body as the source.

What it means. A draft is not something you can hold a supplier to. A finished specification is, so a seller offering to run your model inside its auction can now be asked which version it builds to and what it does when the answer is no. The addition worth knowing is that one container may call another, including one run by a different company, so the thing deciding inside the auction is no longer necessarily a single supplier's.

• your programmatic team If a supply partner offers to run a model inside its auction, does it build to the finished specification or to something of its own, and will it say so in writing?
• your investment team If a container inside our supply partner can call another company's container, whose code has touched our bid, and does any contract we hold name them?

Happened 12 August 2026Published 24 September 202643 days later

Touches 2 modules across 1 book.