Meta made Muse, its personal AI agent, available in Australia on November 17, 2026, following its global announcement, according to the Meta Newsroom. Muse runs on iOS, Android, the web at muse.ai, and Mac. It compares products and supports purchases, but the user must approve before it buys anything. Shopping and services connectors are rolling out progressively with nine named early Australian partners, including Kogan, MECCA, Officeworks, ShopBack, Canva, and Ticketek. The same week brought OpenAI's Businesses in ChatGPT on November 10, Google's expanded visible AI labels announced November 12, and ChatGPT Wallet on November 13.
Meta's announcement does not describe ad placements inside Muse, and Kopa does not run Meta campaigns. Its role is the channel you control directly, which is Google: it keeps your Merchant Center feed accurate, and as an AI Google Ads media buyer it captures live search demand, blocks wasted queries, and executes only what you approve.
For store owners, this week's work comes down to three checks. Your product data needs to be accurate everywhere, your product information needs to hold up in a side-by-side comparison, and you need a way to see orders an agent places.
Agent shopping updates at a glance
| Update | Effective date | Default behavior | Risk for online stores | What to do |
|---|---|---|---|---|
| Meta Muse in Australia | Nov 17, 2026 | Compares products, purchase needs user approval | Weak product data loses comparisons | Audit titles, prices, stock |
| ChatGPT Wallet | Nov 13, 2026 | User approves, agent pays with a one-time token | Orders arrive with unclear source | Baseline attribution, review fraud rules |
| Businesses in ChatGPT | Nov 10, 2026 | Verified profiles add feeds in beta, Business Agents US first | Feeds drift out of sync | Keep one source of product truth |
| Google visible AI labels | Starting Nov 2026 | Labels may show on AI creative | Disclosure surprises in new regions | Log which ads use AI creative |
- Businesses in ChatGPT
Verified profiles upload feeds, US agents
- Google AI labels expand
California, South Korea, Vietnam
- ChatGPT Wallet
User approves, agent pays with a token
- Meta Muse in Australia
Compares products, purchases need approval
- Your move
Audit product data and attribution
1. Meta Muse arrives in Australia with shopping connectors
Muse is a general-purpose agent. It does not live inside a single store or marketplace. Per Meta's announcement, it connects to retailer and service partners, compares products, and supports purchases that the user approves first. Australia is the latest expansion market, and the local connectors are rolling out progressively rather than all at once.
This matters because it puts a layer between the shopper and your store. When a shopper asks an agent to compare options, the agent works from whatever product information it can reach. Meta's post names the partners but does not detail what data each connector exposes. Plan for the cautious case: your product facts get read, summarized, and weighed against a competitor's without anyone visiting your product page.
What to do this week:
- Check whether you sell through any named connector partner, such as Kogan, Officeworks, MECCA, or ShopBack. If you do, the product data you send that partner may be what Muse sees.
- Compare your product titles, prices, and stock status on those partners against your own store. Mismatches that a shopper would forgive can sink you in an agent's side-by-side comparison.
- Rewrite vague titles so the key attributes are explicit: size, material, compatibility, quantity, and model. An agent comparing three kettles needs the capacity written in the data.
- Make sure each variant has its own accurate price and availability. Agents that support purchases need to know exactly which variant is in stock.
2. ChatGPT Wallet makes approval-gated agent payments a pattern
OpenAI introduced ChatGPT Wallet for purchases made by ChatGPT agents. The user approves the purchase with biometrics, and the agent completes checkout with a one-time network payment token, without access to the buyer's actual card details. OpenAI says Wallet is rolling out first to US ChatGPT Pro subscribers aged 18 and over.
Muse's approval requirement and Wallet's approval step point the same way. Platforms are converging on an agent that does the legwork and a human who confirms the spend.
For merchants, the open question is measurement. Neither announcement explains how agent-completed orders will appear in your analytics or ad platforms. Some of these orders may arrive with little or no campaign context, which would show up as direct or unattributed revenue.
What to do:
- Record your current share of direct and unattributed orders now, before agent purchases grow. You need a baseline to spot a shift.
- Review fraud and bot rules on checkout. A tokenized agent payment can look different from a human checkout, and you do not want to block approved buyers.
- Confirm your conversion tracking is clean so you can tell real channel changes from tracking gaps. Kopa's conversion tracking guardrails audit this continuously on the Google Ads side.
- Brief customer service on orders where the shopper approved a purchase their agent assembled. Returns and disputes may reference the agent's summary rather than your product page.
3. Businesses in ChatGPT asks merchants for product feeds
On November 10, OpenAI introduced Businesses in ChatGPT. In beta, eligible businesses located where ChatGPT is available can claim a profile in Business Hub and, once verified, add self-service product feeds with pricing and availability. Business Agents, which can help a shopper compare two products using the business's own information, initially require an eligible US ecommerce or online-service business.
Businesses can sponsor those agents as destinations for ChatGPT Ads. Sponsored placements are labeled and kept separate from ChatGPT's responses.
For Australian stores, the feed side may be open now, subject to verification, even though agents start in the US. Either way, it adds another product feed to maintain, and every new feed is another place where a price or stock status can go stale.
What to do:
- Pick one source of product truth, usually your store platform, and push to every feed from it.
- Treat your Google Merchant Center feed as the reference standard. If it is accurate, the same data is ready for agent surfaces as they open in your market.
- Schedule a weekly spot check of ten best sellers across every feed you run, comparing price, availability, and title.
4. Google expands visible AI labels on ad creative
Google said its visible AI-labeling functionality would expand starting November 2026 to regions including California, South Korea, and Vietnam, per its AdSense Help announcement. Labels may appear on creatives served through Google demand, including Google Ads and Display & Video 360.
This sits alongside the agent news for a simple reason: as AI shows up on both sides of the shopping journey, platforms are making its role visible. If you generate product images or video with AI, expect disclosure to become more common.
What to do:
- Keep a log of which live ads use AI-generated or AI-edited creative.
- Check whether any of those ads serve in the named regions, and review how a visible label would read next to your offer.
- Keep generated product visuals faithful to the real product, so a label never sits on an image that misrepresents what ships.
5. Google Search remains the demand you can measure directly
Agent shopping is real, but the Muse and Wallet announcements give advertisers no reporting, bidding, or placement controls to plan against. Google Search is different. A shopper who types a product query has stated intent, and you control the keywords, the ads, the landing page, and the conversion tracking.
That makes Search the practical counterweight while agent channels mature. If agents take some comparison shopping off your product pages, the buyers who still search need to land on accurate pages with clean tracking, and your budget should not leak to irrelevant queries. Kopa verifies real buyer demand with live search volume, writes ads grounded in your live pages, pauses out-of-stock landing pages, and helps Meta-first brands with adding Google demand beyond Meta ads.
What to do:
- Audit your top Search and Shopping campaigns for wasted queries before holiday traffic peaks.
- Make sure ads never point to out-of-stock or broken pages, since agents and shoppers both punish dead ends.
- Compare Search conversion rates month over month. If agent shopping grows, a shift in direct traffic alongside stable Search performance tells you the change is happening outside paid media.
What good looks like
- Product data matches everywhere: titles, prices, and stock agree across your store, Merchant Center, and any connector partner such as Kogan or Officeworks.
- Attributes are explicit: every product title and description states size, material, compatibility, and quantity in plain words an agent can compare.
- Feeds have one source: any new feed, including a Business Hub feed, pulls from the same product data as Merchant Center.
- Attribution has a baseline: you know your current direct and unattributed order share and review it monthly.
- Checkout accepts approved agents: fraud rules are reviewed so approved, tokenized agent purchases are not blocked.
- AI creative is logged: you know which ads could carry a visible AI label and in which regions.
- Search demand is protected: wasted queries are blocked, out-of-stock pages are paused, and conversion tracking is verified.
You can get started for free at kopa.ai.
