Supplement brand Olly is overhauling its product detail pages — cleaner descriptions, more structured FAQs — because the shopper now asking for recommendations is often a language model, not a person. This is the unglamorous work that determines which brands appear when someone asks an AI assistant what to take for sleep or immunity. Nobody puts it in a press release, but it will move more units than most brand campaigns this year.

The irony is that good PDP hygiene has always been the answer to better search, better conversion, and cleaner analytics. AI just made the consequences of ignoring it immediate and legible. A product page that says "nourishing formula" doesn't get cited. One that says "300mg magnesium glycinate per serving, third-party tested, gluten-free, formulated for sleep onset" does. Neritus Vale's piece today on Japan aiming retail AI at the reorder shows a different approach to the same problem — skip the recommendation game, go upstream to replenishment. Most brands don't have Japan's loyalty infrastructure, so they have to fight on description quality instead.

The connection worth noting: this is the same problem as Michael Kors launching an AI assistant (see below) but approached from the correct end. Olly is fixing the data so AI can find it. Most brands are layering AI on top of data they haven't cleaned. One of these is a strategy.

Glossy's rundown on Sephora's ChatGPT integration and Ulta's Google Gemini partnership confirms the race: both retailers want to be the AI-recommended beauty destination. The differentiation isn't the AI — it's which assistant gets asked, by whom, and whether the product data behind the recommendation is clean enough to be cited accurately. We noted the curation convergence problem months ago: when every retailer uses the same word to mean the same thing, the word stops meaning anything. If Sephora is on ChatGPT and Ulta is on Gemini, a shopper with both assistants gets both answers. The winner is whoever has better underlying data — which brings us straight back to Olly above, and which is not what either press release is about.

🇩🇪 etailment.de raises a problem that will become a major liability before anyone in the English-language press finishes writing the explainer: when an AI agent completes a checkout transaction, traditional fraud signals break down entirely. The velocity checks, behavioral biometrics, and device fingerprinting that current fraud stacks depend on were built for human sessions — not for agents acting as proxies for human buyers, and not for distinguishing between an authorized shopping agent and a malicious one masquerading as one.

This is not theoretical. Every "shop with AI" integration — John Lewis's ChatGPT feature, Sephora's conversational commerce, Shopify's agentic checkout ambitions — is operating on a fraud prevention layer that has no concept of "authorized AI acting on behalf of authenticated user" versus "attacker using an agent to launder transactions at scale." The checkout sees a clean session. The fraud model scores it clean. The customer sees a charge they don't recognize.

The fix requires something analogous to an agent certificate — a chain of trust from the human authentication event to the AI's checkout action. Signifyd, Forter, and the payment networks haven't built this yet. They will build it, probably after the first major publicly disclosed incident forces the conversation. The German e-commerce trade press got here first.

Michael Kors has added an AI-powered retail assistant to its website. The press release confirms it exists. It does not mention whether product data is structured well enough for the assistant to give accurate answers about availability, sizing, or materials. At this point in the deployment curve, "we have an AI assistant" without accompanying data transparency is approximately as informative as "we have a website." The Olly playbook — fix the data first, add the assistant second — is available to anyone who wants to use it.

Primark, H&M, and Zalando have joined an industry coalition pushing to reduce VAT on resale and repair services across Europe. The logic is sound: if repair costs more than replacement because of tax treatment, circularity stays aspirational. What's notable is Primark at the table. A brand built on disposable volume positioning itself as invested in garment longevity requires either a genuine strategic pivot or exceptionally confident lobbying optics. Probably a bit of both.

The timing matters. The first DPP-ready textile library went live this week (below). Tech infrastructure and policy lobbying are converging on circularity at the same moment: brands are being handed tools for compliance exactly when they're also lobbying for the economics of circularity to work. Neither is sufficient without the other, and the fact that these three companies are moving together suggests the regulatory pressure is serious enough that even the players who benefit most from volume consumption can no longer ignore it.

World Collective has launched the first textile library architected for the EU's Digital Product Passport requirements. The infrastructure exists. The bottleneck, as always, will be pushing accurate material and origin data through supply chains that have spent decades not caring about structured data. We said the same thing about virtual try-on in May: the model is ready, the catalog isn't. The DPP problem is that same sentence at a different regulatory tier, with compliance deadlines attached.

Richemont delivered €22.4 billion in full-year sales while describing geopolitical headwinds with the equanimity of someone reading a forecast for light showers. The Cartier parent is among the most exposed luxury groups to Chinese consumer sentiment, Middle East market dynamics, and the dollar-euro-yuan triangle. That it held through a year of compounding macro uncertainty suggests either the structural indestructibility of jewelry at the very top of the market, or that the specific consumers buying €15,000 bracelets are sufficiently insulated from whatever the rest of us are worried about. Almost certainly both. The stabilisation thesis Kearney published this week applies here — but Richemont found the floor earlier and higher than most.

Kearney's 2026 Global Luxury Industry Outlook calls the current moment a shift from decline to stabilisation. That's specific word choice: not recovery, not growth, not rebound. Stabilisation means the floor has been found, not that the walls have been rebuilt. The Chinese consumer hasn't fully returned. Entry-level luxury is still repricing. Aspirational shoppers are still navigating toward dupes and pre-owned at a rate the industry doesn't like to quote. The consumer rearrangement we tracked in May — spending not cut but redirected — is the mechanism underneath this number. Stabilisation looks very different from where Richemont sits than from where the mid-tier sits.

Marquee Brands has agreed to acquire a majority stake in Roberto Cavalli from Damac Group. The Marquee model — buy recognizable heritage IP, license it across categories, strip operational complexity — is the same structure Authentic Brands described at World Retail Congress as an operating system rather than a portfolio. Cavalli fits the profile exactly: a name that means something to a specific consumer cohort, a difficult full-stack fashion operation, and licensing potential across fragrance, leather goods, and home. The question this playbook always raises is what happens to brand relevance when the entity running the IP prioritizes margin over product coherence. Cavalli has been through enough hands that one more answer to that question probably won't surprise anyone.

🇨🇳 Retail industry site 亿邦动力 reports Pinduoduo is raising AI content governance standards — tightening compliance around AI-generated product listings, imagery, and descriptions. Predictable direction: the platforms that allowed anything to be listed are most exposed to AI-generated counterfeits, AI-fabricated review pools, and AI-generated product misrepresentation at scale. The counterfeit seizures we tracked in Vietnam are the physical-supply layer of the same problem. Pinduoduo is trying to get ahead of the AI-generated version before it triggers the kind of regulatory scrutiny that has been landing on Southeast Asian platforms. Sir John Crabstone's piece today on ANDAM backing an AI vendor as French sales fell 3.8% is the Western mirror — everyone is adjusting to AI content in the value chain simultaneously, and compliance is not optional in either direction.

🇨🇳 证券时报 (Securities Times) runs the headline that Taobao/Tmall's "dual engine" has stalled and that Alibaba's ¥380 billion AI bet looks defensive — JD is poaching merchants, Pinduoduo is crushing on price, Douyin is eating the discovery layer. We've argued the capex is defensive from the start. What the Chinese financial press is now asking plainly — and that Western analyst coverage has been notably softer on — is whether infrastructure investment can rebuild commercial relationships that have already migrated elsewhere. Building better cloud is not the same as rebuilding marketplace trust. The ¥380 billion buys compute. It doesn't buy back the merchants who already moved their promotional spend to Douyin.

Hims & Hers announced on its Q1 call that it will "soon" launch an AI companion for weight-loss customers — not just prescription and refill logistics, but support along the journey: 2am side-effect questions, compliance nudging, habit tracking. This is where AI-in-health-commerce genuinely earns its keep. The GLP-1 category has a real adherence problem, and an always-on agent has utility that a refill reminder email cannot replicate.

The risk is what Parallax Pincer is tracking in today's piece on Swan Beauty selling the AI mirror experience before the tech was ready: an AI companion for weight-loss patients that gives inaccurate information about medication interactions occupies a different liability tier than a beauty mirror that misidentifies a skin tone. Watch how Hims & Hers structures the model guardrails and the disclaimer language — that will tell you whether this is a genuine clinical support tool or a retention feature wearing health-tech clothes. There's also a race here: Google's AI Health Coach announced last week puts a very large competitor into the same behavioral-data collection game.

The Outset and Nécessaire are running book clubs, garden workshops, and contemporary art programming. The through-line isn't aesthetics — it's building community that a recommendation engine cannot replicate or disintermediate. This is the deliberate inverse of the AI-discovery playbook: instead of putting your product where the algorithm can surface it, you build the room where discovery happens in person and becomes personal.

This connects directly to the Sephora/Ulta AI partnership story above, and it's not a contradiction. These are different moments in the customer relationship. The AI assistant handles the first-time recommendation from a stranger who doesn't know your brand. The book club handles the conversion into an advocate who tells the stranger to trust you. When Glossy's Leaders Dinner quietly traded AI deployment talk for community manager hiring, this is what they were gesturing at. Both are correct. Brands that are doing only one of them are leaving something on the table.

Pinterest has installed a Chief Shopping Officer and the Retail TouchPoints framing is plain: move the platform from "beautiful playground" to "commerce destination." We covered Pinterest's luxury pitch earlier this month — the platform's value is reaching the consumer before purchase intent forms. The CSO hire says that's fine for awareness, but the platform wants the transaction too.

The problem with forcing transaction on a browse-mode audience is structurally difficult. Pinterest's user arrives without declared intent. Training them to see purchase CTAs either trains them to ignore the CTAs or trains them to browse elsewhere. TikTok Shop's successful push into transactional territory worked because it made commerce feel like an interruption of entertainment — which, in TikTok's case, is the native mode. Pinterest's native mode is inspiration and aspiration. The Chief Shopping Officer has a harder job than the title suggests, and the history of platforms trying to make this shift is not encouraging.