A $660 perfume started getting duped on TikTok — same vibe, $18 — and Guerlain, which has been making fragrances since 1828 and has created over 1,100 of them, finally decided to pay a creator. That is the actual story. Not "heritage brand enters the creator economy" but "the dupe conversation forced a decision that 198 years of prestige positioning never required." Guerlain's first paid influencer campaign is a reaction to being placed in a direct comparison it previously didn't have to have.

This is the mirror image of today's piece on Garnier, which spent into culture infrastructure before the dupe conversation arrived. Garnier bought the moat. Guerlain is filling it after the breach. Both can work. Only one is on offense.

The measurement problem will be instructive. Guerlain enters influencer marketing with no internal baseline, no historical ROI benchmark, and an audience that may or may not respond to paid creator content from a brand whose credibility rested precisely on its remove from that world. The first campaign is also the first test. Today's piece on book clubs as beauty's hedge against the algorithm is relevant here: the brands building off-platform community aren't reacting to dupes — they're structuring against them before the comparison arrives.

Richemont delivered €22.4 billion for the fiscal year through March 2026, described in-house as "solid performance amidst geopolitical shifts." Cartier hard goods remain apparently recession-proof and apparently war-proof. We argued in May that luxury's pricing premium survives macro anxiety better than consensus expects — Richemont's numbers are consistent with that thesis. The phrase "geopolitical shifts" is doing real work here, though. It pre-positions a qualifier for the next downgrade announcement without marking the current quarter as anything other than a win. File under: alibis drafted in advance.

Pepco Group — pan-European value apparel and general merchandise — posted H1 revenue up 5%, profit after tax up 52.3%. In the same fortnight, Richemont posted €22.4 billion. Both companies are reporting strong numbers. The barbell is structurally intact: ultra-luxury holds, deep value expands, and the middle of the price spectrum publishes turnaround plans and calls them strategies. Pepco gets no analyst keynote, no Glossy feature, no LVMH comparison. It is simply making money at scale while everyone else debates positioning.

Marquee Brands — the company that manages Martha Stewart, Ben Sherman, and the Marilyn Monroe estate — is acquiring a majority stake in Roberto Cavalli from Dubai's Damac Group. This is not a fashion investment. It is an IP extraction. Marquee doesn't manufacture. It licenses. Cavalli becomes a name on handbags and fragrances that Marquee doesn't touch but collects royalties from.

Today's piece on licensed goods outgrowing retail covers the structural logic: when the IP is the product, the physical supply chain is someone else's problem entirely. Cavalli has all the elements that make an IP play viable — visual signature with instant recognition, prestige recall that survives years of troubled ownership, and a discount entry price courtesy of that troubled ownership. The next two Marquee announcements will be a handbag licensee and a fragrance deal. That's the playbook. It will work, which is the infuriating part for anyone who thought Roberto Cavalli was a fashion company.

🇩🇪 German source, translation mine. This etailment.de piece asks the question the agentic commerce press releases omit: when an AI agent completes a checkout on a human's behalf, how does fraud detection distinguish a legitimate agent from a well-scripted attacker? Current fraud models are calibrated on human behavioral signals — typing cadence, mouse hesitation, the micro-pauses of decision-making. An authorized AI agent executing a pre-confirmed purchase looks cleaner and faster than any human, which is exactly how a sophisticated fraud script also behaves. The signal is the same in both cases.

The retailers who've been shipping agentic integrations into ChatGPT and Claude faster than the benchmarks can keep up haven't solved this. The fraud vendors haven't either, at least not publicly. The gap is between the demo checkout experience — seamless, delightful — and the authorization model underneath it, which was designed for a human sitting at a keyboard. Worth watching: whether Signifyd, Forter, or Riskified files anything on agent-specific authentication frameworks in the next quarter. Until they do, every agentic checkout integration is an open question at the authorization layer.

🇨🇳 Chinese source. Today's full piece covers this in depth, but the comparative signal is worth flagging separately: Pinduoduo is raising the compliance bar on AI-generated content at exactly the moment Western platforms are still treating the problem as theoretical. This is a pattern worth tracking. Chinese e-commerce platforms have consistently built enforcement infrastructure ahead of Western equivalents on fake reviews, counterfeit listings, and pricing manipulation. The trade-press framing is always that China is catching up to Western standards on transparency and consumer protection. On platform compliance mechanics, the direction often runs the other way.

🇨🇳 Chinese source. Securities Times analyzes Alibaba's Taobao "dual engine" — Taobao for small merchants, Tmall for brands — and finds it decelerating as Pinduoduo absorbs price-sensitive buyers, Douyin takes entertainment-led discovery, and JD holds logistics-dependent categories. The ¥380 billion AI infrastructure pledge, which we covered when it was announced, reads here as defensive: excellent infrastructure for a platform that customers have already learned to route around for specific purchase occasions. Infrastructure solves latency and reduces cost. It does not dissolve the habits consumers formed at Pinduoduo's better prices or in Douyin's better entertainment. The pipes are not the problem.

Hims & Hers announced an AI companion for weight-loss users during Q1 earnings. A week earlier, Google launched an AI Health Coach through Alphabet. Different entry points — Hims has the pharmacy and the existing patient relationship, Google has the search query history, Fitbit, and seventeen years of health-adjacent behavioral data — but the same underlying bet: the entity that owns longitudinal health behavior data owns the most powerful purchase-intent signal ever assembled in consumer goods.

The commerce implications are underreported. A health behavior graph that knows what you eat, how you sleep, what medications you take, and how your weight fluctuates can predict supplement purchases, apparel sizing events, and gym membership patterns with an accuracy no standard personalization model approaches. The memory layer argument we made in May applies here at maximum scale: longitudinal behavioral data wins the next round of personalization, and both companies are building for exactly that layer. The question no one in the press release asked is what consumers understand themselves to be trading when they open the app.

World Collective has launched what it describes as the first Digital Product Passport-ready textile library — fabric and material data pre-structured for the EU's incoming DPP requirements. Today's piece argues that the DPP turns sourcing into a data contest; this is the first piece of shared infrastructure that makes the contest joinable for brands that lack the resources to build proprietary cataloguing systems. Brands that populate this library before the regulation goes live have a structural head start. Brands still tracking materials in spreadsheets will spend their compliance budget catching up instead of competing on the content of their passports.

Michael Kors has added an AI retail assistant to its website, joining roughly every mid-to-large fashion retailer that has announced the same capability in the past eighteen months. The assistant helps with product discovery. What none of these deployments address is the second-order problem: when a customer's shopping agent arrives on behalf of the human, the onsite assistant is working for the brand and the visiting agent is working for the customer. These are not aligned goals. The brand's assistant wants a conversion at full price. The customer's agent wants the best available price across all channels, including competitors. Agentic shopping benchmarks suggest this collision is closer than most retail AI deployments are structured to handle.

Glossy checked in on Sephora's ChatGPT integration and Ulta's Google Gemini partnership — both real, both live, both solving for the same problem: product discovery in a catalog that has outgrown human curation. As we noted in May, every major beauty retailer is pitching "curation" as its competitive differentiator. Adding AI doesn't resolve the undifferentiation — it commoditizes it at speed. The actual competitive advantage lives in the proprietary data the AI runs on. Neither company disclosed what that is.

Alana Pallister sold her house to fund 300,000 units of the Blare tracksuit. She sold one million in under five months. The I.AM.GIA story is a case study in conviction inventory: the bet that a single viral item produced at confidence scale beats a diversified SKU portfolio hedged against uncertainty. She was right this time. The Coachella follow-up is the harder test.

Brands that built durable businesses post-viral typically did it with a slower, quieter, more boring second act — not a second viral moment. The ones that chased the moment found that the first moment had consumed the authenticity that made it possible. Today's piece on Reale Actives covers a different version of the same pressure: what happens when you sell out and the backlash comes anyway, and you spent three months pre-writing the defence. The I.AM.GIA test will arrive when the third product doesn't go viral. That's when we find out what the brand actually is.

Kantar's 2026 fashion outlook uses "treatonomics" to describe a consumer who isn't spending broadly but is spending specifically — constrained baseline, concentrated and emotionally-driven splurge. The word is ugly. The observation is accurate. The consumer who can't or won't spend on basics will still buy the limited drop, the event-adjacent purchase, the thing that feels earned. I.AM.GIA's tracksuit is a treatonomics purchase. Guerlain's $660 perfume is a treatonomics purchase. The brands winning at the margin this year are the ones positioned at the treat occasion rather than the staple occasion — and most of them arrived there by instinct, not by reading the Kantar deck.