Log entry Nº 692 · the stack watch

Bearing S 180° · November 17, 2026 · 5 min read

Commerce TechS 180°Watch log

Resale Platforms Rewire Discovery for the Age of AI Shopping Agents

Vestiaire feeds listings to OpenAI agents, The RealReal restructures 1.5M items for AI discovery, and Poshmark claims five-times faster listing as resale rewires search.

By Rebecca Stone5 min read924 words

How To Crack Resale’s AI Algorithms
Fix Nº 692How To Crack Resale’s AI Algorithms — LUM3N / pixabay

Ask ChatGPT for a designer handbag and the answer may soon be a pre-owned one. Vestiaire Collective is preparing to feed its inventory into OpenAI's broader agentic shopping systems, so secondhand pieces appear alongside new products in AI-generated recommendations. The move follows a quieter step: last month the resale platform launched a dedicated app plugin inside ChatGPT, letting shoppers search its pre-owned luxury stock and compare products through conversational prompts.

Rémi Bouchez, appointed Vestiaire Collective's CTO in March, frames the shift as structural. A shopper hunting a new Prada bag could be shown a pre-owned version at a fraction of the price, or an archival design no longer available firsthand. If AI agents such as Muse, Instinct and Duet reach mainstream adoption, the way luxury shoppers weigh options could change fundamentally.

"These agents are capable of ingesting a ton of information — way more than human shoppers can," Bouchez says. "They remember user preferences over time, and they can catch things like discreet terms and conditions or shady seller behaviors and put them in front of a user who would have otherwise missed them." An agent could compare a Prada bag across platforms, assessing condition, authentication details and seller reliability before recommending the best value option.

What are rivals doing?

The RealReal introduced its own AI shopping assistant, Ask TRR, last month. Built on Google Gemini AI, it supports natural-language search and image upload for visual search, and learns shopper preferences. On the company's third-quarter earnings call last week, CEO Rati Levesque said Ask TRR's early adopters spend 20% more per transaction than non-users. The RealReal is also restructuring data on its 1.5 million listed items so external AI agents can find and recommend them more easily.

Depop, meanwhile, has enabled shoppers to search its marketplace through Apple's Visual Intelligence on supported iPhones.

How does discovery get more visual?

At Vestiaire Collective, visual search now recognises multiple objects in an image and identifies similar listings, while search systems increasingly interpret natural language. AI also extracts information from photographs — aspects of fit, visible defects — that previously depended on sellers filling in descriptions.

Poshmark's "Smart List AI" generates searchable titles, descriptions and categories from uploaded photos. Chief product officer Heather Friedland says listing on Poshmark is now five times faster, and automated offers let Poshmark negotiate with buyers on the seller's behalf, making selling 50% faster.

"We want sellers to focus most on what they do best: source great items and have Poshmark take care of the rest," Friedland says.

At Depop, more shoppers browse the personalised Suggested For You feed than start with a search, says chief product officer Sonia Biddle. Its AI identifies listings by category, colour and style, making products discoverable even without exact keywords. "Fashion often starts with an idea: an era, an occasion or someone whose style you love, and that's not always something you can put into words," Biddle says. "AI helps us understand what you're after, not just the words you type."

Last month Vestiaire announced a partnership with AI photography company Photoroom to standardise the main photograph of every new listing — around 40,000 items a day. The technology removes backgrounds, adjusts lighting, centres products and adds realistic shadows. A randomised A/B test found processed listings had a 5% higher sell-through rate than those using Vestiaire's previous in-house tool, with unique product page views up 3% to 7%. Bouchez says Vestiaire has seen roughly a 6% improvement in the share of enhanced listings selling within seven days, without a corresponding rise in returns.

Next come more ambitious enhancements, such as "digital ironing" — though an algorithm that erases a scuff could undermine the trust resale depends on.

How can sellers rank higher?

Vestiaire uses a three-stage ranking system: identify relevant products, reorder them by marketplace factors, then adjust for individual preferences. Factors include condition, pricing, location and seller reliability; a nearby item may surface first because it means faster shipping, fewer customs complications and easier returns. An implausibly cheap luxury item may raise authenticity questions rather than look like a bargain. "Trust is a big pillar for Vestiaire — we invest a lot in that trust, on manual verification, on assessing a listing before it's published," Bouchez says. "We want to keep the same level of trust but lower the investment with AI."

Depop's algorithm prioritises query match — brand, category, size, colour — plus popularity signals such as views, likes and bag additions, alongside seller history and location. For underperforming listings, Biddle recommends new photographs, a more detailed description and a pricing review.

What about personalisation and its limits?

Shoppers are deliberately "training their algorithm," Biddle says, curating what they like, view and save. To counter hyper-personalisation, Poshmark launched a "Tastemaker" programme in September, enlisting fashion figures including Courtney Grow, Colleen Allen and Jalil Johnson to curate shoppable edits, backed by a dedicated human merchandising team.

"On trends, a purely algorithmic system tends to amplify what's already popular," Friedland says. "Human curation is our counterweight."

For brands, resale data now signals enduring demand. "Resale tells brands something their own sales data can't: what people still love years after it left the shop floor," Biddle says. Poshmark uses AI to track demand for designers, eras, silhouettes and price points. If AI assistants connect consumers with archival products as readily as current collections, brands may compete with their own past successes — without controlling how those products appear.

Plot points

  1. Vestiaire Collective launched a ChatGPT app plugin last month and is opening its inventory to OpenAI's agentic shopping systems.
  2. The RealReal's Ask TRR users spend 20% more per transaction; the company is restructuring data on 1.5 million listings for external AI agents.
  3. Poshmark says Smart List AI makes listing five times faster and selling 50% faster via automated offers.
  4. Vestiaire's Photoroom partnership processes ~40,000 listing photos daily; A/B tests showed 5% higher sell-through and 3-7% more page views.
  5. Poshmark launched its Tastemaker curation programme in September with human merchandisers countering algorithmic trend amplification.

via vogue.com (Source)

Filed under

  • ai-shopping-agents
  • resale-marketplaces
  • visual-search
  • personalization
  • luxury-resale
Share this article:

More from Rebecca Stone

Rebecca Stone

Show full bio

Staff writer covering companies, deals and strategy in marketing technology at Martech Beacon.

38 articles

Neighbouring fixes

‹ Previous article