What has my attention, right now.

The OAIC has been busy. It has published new guidance, fact sheets and a flowchart on AI and automated decision-making transparency, ahead of obligations commencing on 10 December 2026.

The requirements concern privacy-policy disclosures where personal information is used by a computer program to make, or substantially and directly contribute to, decisions that could significantly affect someone’s rights or interests.

For marketing teams, I’d start with lead scoring, offer eligibility and personalised pricing, then check which decisions meet that threshold. The commercial question is what the system changes for the customer, and whether the business can explain it. Buying the model from a martech vendor still leaves that question to answer.

Read the OAIC’s announcement and resources.

Advisory, 7 October: Automated decisions: what marketers need to do.

The government has released the Digital Duty of Care exposure draft, proposing stronger safety responsibilities for online services. Its “My Feed, My Way” proposal would give social media users a choice over personalised recommendations in their default feed.

My read: brands relying on recommendation-led discovery should watch this closely. Greater choice over feeds could change organic reach and the balance between creator partnerships, paid media and owned audiences. The way platforms distribute attention is becoming a regulatory issue with commercial consequences.

This is proposed online-safety legislation. Feedback on the draft closes at 12pm on 22 September 2026.

Read the exposure draft.

Marketing scopes are to-do lists in costume. A list of tactics is not a plan.

What I want instead: every tactic tied to a specific performance lever, with the outcome named and the measurement defined before anyone starts. If a tactic cannot point to a lever, it does not make the list. If a lever cannot point to a client outcome, the lever is wrong.

Static by default, with interactivity where it earns its place. The islands pattern is the right primitive for the content sites I am building.

The full argument, including the distinction between build-time rendering and edge serving, is in Why static-first architecture matters more than schema.

Websites do not disappear. They become backend infrastructure.

By 2030 the transactional layer of the web is agents negotiating with brand APIs. The question worth asking now is not “how does my site rank” but “what does my site expose, and to whom”.

THE ICONIC’s help page, dated 9 July, describes a limited Australian UCP integration: selected own-label products, a single-item purchase, linked accounts and Google Pay. Retailer documentation.

The page gives its start as “Thursday, 10 July 2026”, but that numbered date falls on a Friday. This entry therefore records the documentation date. It does not claim an independently confirmed launch day.

The useful next step is to test the described journey and record the result. Until then, the retailer’s published scope is the evidence available here.

The OAIC has released the draft Children’s Online Privacy Code. Its proposed direct-marketing rules require consent, consistency with the child’s best interests, and personal information collected directly from the child.

For marketers, this goes to how younger audiences are built and reached. My starting point would be the apps, games and websites children actually use, then the data collection and advertising integrations behind them. For services covered by the Code, those choices could change which audience-building and personalisation practices remain workable.

The Code must be finalised and registered by 10 December 2026. The consultation is an opportunity to test the proposals against those workflows.

Read about the Code.

OpenAI says the first version of Instant Checkout did not give merchants the flexibility it wanted. It is putting more emphasis on product discovery while allowing merchants to use their own checkout experiences. OpenAI’s update.

Shopify’s account of the same development describes shoppers completing purchases on a merchant’s store through an in-app browser. Shopify’s announcement.

This is a material change to the shopping journey introduced with ACP in September. It separates the place where a product is discovered from the system that manages the purchase.

The announcement alone does not settle the protocol’s future. It does make checkout flexibility a concrete issue to follow: customer accounts, merchant requirements and service continuity all need somewhere to live.

Google says UCP-powered checkout is rolling out for US shoppers buying from Etsy and Wayfair in AI Mode and Gemini. Shopify, Target and Walmart remain on its coming-soon list. Google’s update.

This moves the record beyond January’s announcement to a named deployment, with named retailers and a stated geography. Google also illustrates a Wayfair journey that progresses from discovery through order review to completion.

The evidence remains a platform report. It does not provide an independent test, a transaction volume or a measure of how often customers need help. Nor does the US rollout establish Australian checkout availability.

For the timeline, the distinction is useful: the protocol has an implementation in a consumer experience. The next questions concern its reach and performance.

Google has announced UCP, the Universal Commerce Protocol, with retail and platform co-developers. The ambition is a shared language across the shopping journey. Checkout for eligible US retailers in AI Mode and Gemini is described as coming soon. Google’s announcement.

The technical starting point is a business describing the services it supports. Google’s accompanying walkthrough shows an agent discovering those capabilities through a standard profile before interacting with the business. Technical introduction.

The announcement creates two histories to follow: development of the protocol, and deployment of experiences that use it. A partner list can tell us who is participating. The later rollout evidence will tell us what a customer can actually do.

Stripe has announced Agentic Commerce Suite, bringing product discovery, checkout and payment integration into one offering. Merchants connect their catalog and choose the agents they want to support, while keeping their existing commerce systems involved. Stripe’s announcement.

The release identifies an implementation problem behind the protocol story. A shared specification still leaves merchants with endpoints, access controls, catalog formats and order handling to connect. Stripe is offering a managed route through that work.

That makes this a product milestone in the ACP ecosystem. The announcement should not be read as proof that every listed merchant is live on every agent, or that the same experience is available in Australia.

The next evidence to look for is concrete deployment: which merchant, which agent, which country and which supported transaction.

OpenAI and Stripe have introduced the Agentic Commerce Protocol alongside Instant Checkout in ChatGPT. The initial experience lets US users buy from US Etsy sellers within a conversation. Shopify merchants are described as coming next. OpenAI’s announcement.

ACP provides a common format for commerce between agents and businesses. The customer confirms a purchase; the merchant remains responsible for processing and fulfilling it. Stripe describes secure, delegated payment credentials as part of that exchange. Stripe’s explanation.

The immediate evidence is a bounded shopping experience. It establishes a starting point for agent-mediated checkout, with the geography and merchant scope specified. It does not establish general Australian availability.

The question to follow is what happens as ordinary shopping requirements arrive: different payment methods, delivery options, customer accounts and the merchant’s existing service commitments.