The Securities Times is running the headline Alibaba's PR team doesn't want translated: Taobao's dual engine is losing speed, the ¥380 billion AI infrastructure pledge looks more like a panic bid than a growth plan, and three separate companies are eating the ground underneath it simultaneously. JD is hiring away talent. Pinduoduo is grinding on price. Douyin is taking the attention. We wrote about this dynamic last week when Alibaba made the pledge, and the Chinese financial press is now reading the Q1 numbers the same way: growth deceleration dressed up as strategic investment.

The phrase in the piece — 增速掉队了 — roughly means "growth fell out of the convoy." That's not analyst euphemism. That's the financial press calling it directly. When the domestic press loses patience with the national champion narrative, the story has usually been true for a while.

Galeries Lafayette's Beijing closure is getting filed as a French luxury casualty in a tough market. That framing misses the point. The companion piece about savvy Chinese consumers shopping smarter is the actual explanation. Galeries Lafayette's customer didn't disappear. She learned to find the same goods for less on Dewu, to read the dupe reviews on Xiaohongshu, to compare prices in real time. The department store format offers curation at a margin she no longer needs to pay for.

The Jacqueline Li anecdote in that piece is worth noting: she was "astonished" to learn of the closure, because she hadn't been in years. Not anger, not loss — just surprise that it was still there. The building outlasted the relevance by some distance. The same structural pressure is what the Gulf retail upgrade ran ahead of McKinsey's prescriptions on: when you digitize the middle class faster than you digitize your own shelves, the shelves close.

Korean tech press is picking this up faster than the English-language retail trades: Amazon is moving to sell its Alexa for Shopping AI stack externally to other retailers. This is not a minor announcement. Amazon built the best-trained in-context product recommendation engine on the planet — billions of purchase signals, a decade of A/B tests, the full Rufus architecture. Now it's offering that as a service to competing retailers. We noted in May that Rufus was living mostly in press releases while ByteDance was publishing model weights on arXiv — this changes that calculus. A retail AI arms race just became a retail AI utility market.

Pascal's piece in etailment (German, worth the translation effort) frames the problem cleanly: current fraud prevention systems rest on a single assumption — a human is at the keyboard. Behavioral biometrics, typing cadence, mouse movement heuristics, the whole stack. When a shopping agent executes a transaction, none of those signals exist. The agent moves with machine precision and machine speed, which looks like a bot attack to every existing fraud system.

So what actually happened when a suspicious order fires? One of three things: a legitimate AI agent placed a legitimate order; an AI agent was hijacked mid-session and placed a malicious one; or a human attacker is hiding behind an AI agent mask precisely because they know the behavioral tells won't fire. The checkout can no longer tell. The piece argues — correctly — that fraud prevention has to move upstream, from the transaction itself to identity verification at the point of agent authorization. That's a fundamentally different product category. It's also probably worth a billion dollars to whoever builds it first.

JD.com is reportedly circling Very Group — the British-Irish online retailer behind Littlewoods — at around £2 billion. The strategic logic is clear once you understand how China's major platforms are rerouting around trade barriers: owning a European-headquartered retailer gives JD an origin point that isn't China. It sidesteps de minimis scrutiny, grants access to established payment rails, customer data, and a logistics network with domestic credibility. Littlewoods has run buy-now-pay-later for decades; JD's Jingdong Baitiao installment product would slot straight in.

The interesting question is what JD does with Very's own-brand fashion and homeware inventory. That's not an acquisition for the brands. That's an acquisition for the pipes. The brands get rationalised within two years.

Olly the supplement brand is updating its product detail pages with clearer descriptions and structured FAQs, specifically to perform better in AI chatbot recommendations. This is the 2026 version of keyword stuffing, except it's actually good practice. The insight is simple: AI assistants don't scroll images, don't infer from lifestyle photography, don't read between the lines of aspirational copy. They extract structured facts. If your PDP says "support your wellness journey" instead of "400mg magnesium glycinate per capsule," Rufus or Perplexity recommends the competitor who wrote the latter.

Retailers are building for AI channels faster than evaluation frameworks can keep up. Olly's move is interesting because it's not a major tech investment — it's a copywriting discipline. The question now is whether brands will bifurcate their PDPs: one version optimised for human visual browsing, one schema-heavy variant for machine ingestion. Expect that to become standard within eighteen months.

Shark Beauty and Lookfantastic are joining Topshop's AI-generated catwalk to add a live social commerce layer — meaning an AI creates the clothes, models wear them on a virtual runway, and viewers can buy the closest real items in Lookfantastic or Topshop's catalog while they watch. This is an actual working prototype of something that's been theorised for three years.

The persistent friction point: the try-on models are ready, the catalog metadata is not. Today's full piece from Parallax Pincer on moving garments is the other half of this problem — a coat needs to move convincingly for live commerce to work, and a still image is a solved problem in a world that's now asking for video. Topshop's setup is early and probably creaky. That's fine. This is the experiment that teaches the industry where the catalog gaps actually live.

Alana Pallister sold her house to buy 300,000 units of a stretchy tracksuit. Less than five months later she had sold 1 million. That's not a fashion business — that's a conviction that the market was wrong about demand, backed by a willingness to bet the house on it literally. The Blare tracksuit was designed to go viral; she simply had the inventory to let virality do its job instead of selling out in hour one and watching imitators fill the gap.

There's a capital structure insight buried here that brands get wrong constantly: selling out is often a failure dressed as success. You proved demand and trained your customer to expect frustration. Pallister solved the problem from the supply side instead of the demand side. That's a different kind of founder intelligence. Now she's betting Coachella. The playbook repeats until it doesn't.

The story that finally forced Guerlain's hand: a $660 perfume went viral, the dupe conversation started, and someone at the house decided that silence was not the correct strategic response to 47 million TikTok views. So they launched a paid creator campaign. Their first ever. In 198 years.

The luxury industry is full of houses that confuse scarcity with mystique and mystery with ignoring the internet. Guerlain has over 1,100 perfumes in its archive and sells around 100 today — the curation has always been intentional. What changed is the customer's willingness to have the conversation in public, and the dupe market's ability to route around the price barrier. The choice is no longer whether to engage with TikTok. It's whether to engage on your own terms or let a girl in a bathroom with a $28 alternative speak for you.

This connects to today's full piece on Everlane and Shein: transparency as a product feature can be copied, but the 1828 founding story cannot. Guerlain should probably be leaning on the latter much harder than they have.

Pinterest hired a Chief Shopping Officer — a title that didn't exist at the company last year — and she's on a press tour explaining that Pinterest is moving from "beautiful playground" to "commerce destination." We've written about Pinterest's genuinely distinctive commerce proposition: they're the only major platform where the buyer is undecided. No declared search intent, no social proof from a creator mid-video — just taste still forming. That's a valuable signal if you can monetise it without destroying it.

The challenge is conversion. Pinterest users save things; they don't always buy things. The new CSO's job is closing that gap without making the platform feel like a storefront, because the moment Pinterest gets aggressive about commerce it becomes Instagram, which lost its aspiration premium years ago. The tightrope is real and there's no obvious model for walking it.

Google's parent Alphabet launched an AI Health Coach in May and is pitching it to the wellness and beauty press as a consumer tool. The product is real. The framing as wellness is the polite version of what it actually is: a longitudinal body-data collection machine that gives Google sleep metrics, exercise patterns, nutritional intake, and symptom tracking for millions of users. That data profile is worth more to pharmaceutical companies, insurers, and — yes — beauty and supplement brands than any advertising relationship Google has sold to date.

The race is three-way: Apple has Watch and Health app with years of lead, Samsung has its ring, Google now has the Coach. Whoever wins sits upstream of beauty and wellness purchasing decisions with data that explains not just what you bought but why your body needed it. For a supplement brand like Olly — who is simultaneously rewriting their PDPs for AI chatbots — that's not a competitor. That's a future distribution platform. Expect partnership announcements within twelve months.

Onton, an AI ecommerce startup, has raised $7.5 million. The round is small by VC standards but tells you something: funding is still moving into AI shopping tools despite the market thinning. The question for Onton is the same question for every entrant in this bracket — evaluation frameworks are still catching up to deployed tools and differentiation from Perplexity Shopping, Google Shopping Graph, and every retailer's own assistant is unclear from the outside. $7.5M buys eighteen months. That's probably enough time to find out if there's a real wedge or not.

The new Dior store in Osaka's Shinsaibashi district by Sou Fujimoto features a ribbed concrete facade designed to evoke flowing drapery — the exterior is literally made to look like fabric. Peter Marino handles the interior, which is the usual Marino: theatrical, maximalist in the right ratios. The architecture says something a paid post cannot: this is a permanent place, not a pop-up, not a collab. It's also a useful counterargument to every "retail is dead" think piece. Luxury spent real money pouring concrete to look like Dior silk in Osaka, because the physical store still communicates permanence in a way no app can replicate. Worth a walk-past if you're anywhere near Shinsaibashi.