Connect with us

NEWS

Vestiaire Opens Its Catalog to ChatGPT Shopping Agents

Vestiaire Collective and The RealReal are feeding catalogs to ChatGPT-style agents that shortlist a few bags and leave most listings unseen.

Published

on

Vestiaire Collective is pushing pre-owned listings into ChatGPT so a shopping agent can put a used bag next to a new one. Rémi Bouchez, named chief technology officer in March 2026, said the platform wants that catalog inside OpenAI’s wider agent systems.

A dedicated ChatGPT app went live in October 2026. Sellers still fuss over in-app rank, but the fight that will decide who gets seen is moving off the marketplace.

Vestiaire Is Opening Its Closet to ChatGPT

Bouchez said Vestiaire is working to make product listings available to OpenAI’s broader agentic shopping systems. Inside the October app, a shopper can already search pre-owned luxury with conversational prompts and compare products without opening a dozen tabs.

A person who asked for a new Prada bag can be shown a used one at a lower price, or an archival style that no longer sits in a boutique. Vestiaire’s plugin still compares up to three selected listings and asks the agent to explain gaps in price, condition and seller information. That cap is the new shelf.

AGENT-READY CATALOG AT A GLANCE

  • Daily intake: About 40,000 new Vestiaire listings a day now get a standardised main photo.
  • Agent crawl: Vestiaire bulk-optimised 12 million URLs for AI agents and now serves pre-rendered pages to bots such as ChatGPT-User.
  • Shortlist: The ChatGPT app compares up to three listings, then hands the shopper a reasoned pick.
  • Server load: SpeedWorkers handled 8.76 billion requests over the past year and passed 1.33 billion through to Vestiaire’s own servers.

Botify’s case study also says Vestiaire is preparing tests of product-feed enrichment and AI shopping protocol compliance, which is the unglamorous work of being readable to a machine that never scrolls. On November 10, 2026, OpenAI posted a businesses-in-ChatGPT product, widening the same door.

Muse, Instinct and Duet Already Shop Without the App

Bouchez argues that if agents such as Muse, Instinct and Duet reach mass use, luxury shoppers will stop hopping marketplaces. Those agents, he said, ingest more information than a person will, keep preferences over time, and can surface discreet terms or shady seller behaviour a tired scroller would miss.

These agents are capable of ingesting a ton of information, way more than human shoppers can. 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.

Rémi Bouchez, chief technology officer, Vestiaire Collective

OpenAI is expanding the Agentic Commerce Protocol for discovery, so merchants share product feeds and promotions. Since March 24, 2026, ChatGPT has sent checkout back to the merchant after OpenAI said the first Instant Checkout build did not give stores enough flexibility. Shoppers can browse visually, upload a photo as a prompt, and compare options in one thread. Target, Sephora, Nordstrom, Lowe’s, Best Buy, The Home Depot and Wayfair are already on that discovery feed.

THE AGENTS NAMED IN THE RESALE PITCH

Agent Who built it What it does with a listing
ChatGPT OpenAI Finds and compares products; purchase finishes on the merchant’s checkout
Muse Meta Launched in September 2026; can prepare a live-web purchase for the shopper to approve
Instinct Instinct Invite-only since August 2026; reached by text, call or email
Claude Duet Anthropic Research preview; pays after the shopper approves, including via Stripe Link

An agent that can weigh condition, authentication notes and seller reliability across several apps raises the cost of a sloppy listing. It also means most of the catalog never reaches a human. The person still thinks they shopped. They saw three cards.

A Photo Now Does the Work a Hashtag Used To

Inside the apps, discovery is already sliding from keywords to pictures. Vestiaire’s visual search can recognise several objects in one image and pull similar listings, while search boxes increasingly accept plain language. Software also reads fit and visible defects from photographs, work that used to wait on a seller’s form.

Poshmark’s Smart List AI writes titles, descriptions and categories from the photos a seller already took. Heather Friedland, the company’s chief product officer, said listing is now five times faster, and automated offers make 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,” she said.

Depop’s Sonia Biddle, chief product officer, said more shoppers start in a Suggested For You feed than in a search box. The model reads category, colour and style even when the seller missed the “right” words. Depop also ties into Apple’s Visual Intelligence on supported iPhones, and it lets someone upload a screenshot from a first-hand shop to hunt a cheaper secondhand cousin.

WHAT STILL GETS A LISTING SEEN

  • The pictures: Natural light, several angles, and honest wear now feed both visual search and auto-written titles.
  • The price: An implausibly cheap luxury piece can look fake to Vestiaire’s ranker rather than like a bargain.
  • The seller: Reliability, location and return policy sit beside condition when a global marketplace decides what to show first.
  • The query match: On Depop, brand, category, size and colour still have to line up, then views, likes and bag-adds help.

Hashtag stuffing is the tactic that ages out first. Biddle’s point is simpler than a growth-hack thread: fashion often starts as an era, an occasion, or a person whose style you like, and that is a poor keyword. The image has to carry it.

Listings Edited by Photoroom Sold 5% More Often

Last month Vestiaire said it had partnered with Photoroom to standardise the main photograph on every new listing, about 40,000 items a day, across 70 countries. The system strips backgrounds, fixes lighting, centres the object and adds a realistic shadow, pushing amateur snaps closer to a first-hand product page.

Vestiaire split new listings into two random groups. One used Photoroom; the other used the in-house clipping tool Vestiaire built years ago. Photoroom listings sold at a 5% higher rate of liquidity. Product pages drew 3% to 7% more unique views, with bags and shoes gaining the most. The same images made performance ads about 2% more efficient, and digital authentication was unaffected. The test took eight weeks.

We tested different GenAI solutions, and each time they changed some details. Then it was no longer possible to authenticate from the pictures or to make sure buyers could trust them. Photoroom maintained product fidelity, and that’s clearly the reason we chose it.

Thomas Hezard, chief product officer, Vestiaire Collective

Generic beauty filters already taught the trade a harder lesson. Sellers were running outside AI editors, and return rates went up, because a prettier file is a problem when the object that arrives still has the scuff. Photoroom said that was why Vestiaire switched.

Bouchez said Vestiaire is also testing heavier edits, including digital ironing, to make used pieces look product-page ready. The company has seen about a 6% rise in enhanced listings selling within seven days since those image updates, with no matching jump in returns. That 6% is a later, broader image-update figure, not a restatement of the Photoroom A/B test. The line that cannot move is the defect. If the algorithm wipes a scratch, the authentication story collapses.

What Vestiaire’s Three-Stage Ranking Still Rewards

Until agents take the traffic, sellers still live inside each app’s ranker. Bouchez describes Vestiaire’s system as three stages: find relevant products, reorder them by marketplace factors, then tilt the list toward one shopper. Condition, pricing, location and seller reliability all count. A nearby item can ship faster, dodge customs trouble and come back more easily, so it can surface first.

Fair price is not the lowest number. Vestiaire weighs condition, location and whether a professional seller offers returns. Trust, Bouchez said, is still a mix of people and models, including manual checks before a listing goes live. The aim is to keep that check while spending less of it on human hours.

Depop still ranks on how closely a listing matches the query, then on popularity signals and seller history. If a listing stalls, Biddle’s advice is new photographs, a fuller description and a price review. Shoppers, she said, now talk about “training their algorithm,” liking and saving with intent. The cost of that habit is a feed that stops surprising anyone. She tells people who want out to sort by price or newest, or to use visual search for pieces they cannot name.

Poshmark tried a human counterweight in September 2026 with a Tastemaker programme, bringing in Courtney Grow, Colleen Allen and Jalil Johnson to cut shoppable edits. Friedland said a purely algorithmic system amplifies whatever is already popular, so a merchandising team now sits beside the machine-learning group. Save searches for archival pieces, she said, and like items if you want the feed to learn.

HOW FOUR APPS RANK A LISTING

Platform Main rank signals What a seller can still change
Vestiaire Collective Relevance, then condition, price, location, seller reliability Honest photos, a defensible price, returns if you are a pro
Depop Query match, views, likes, bag-adds, seller history, location New photos, tighter copy, a price cut
Poshmark Photo-derived titles plus likes, saves and tastemaker edits Better photos; saved searches on the buyer side
The RealReal Natural-language and image queries over authenticated stock Consign with complete attributes the agent can read

Those levers still matter on a Tuesday night in the app. They matter less if the shopper never opens it.

Ask TRR Shoppers Spend 20% More per Order

The RealReal spent October wiring the same idea to Google. On October 8, 2026, it expanded Ask TRR to all 45 million members after a pilot, using Gemini Enterprise. Shoppers type a sentence, or upload a picture, and get a shortlist from a catalog in which every piece is one-of-one. Chief executive Rati Levesque said the site holds more than 1.5 million unique pieces every day.

On the October 29, 2026 earnings call covering the quarter ended September 30, she said adoption is still building and that early Ask TRR users spend 20% more per transaction than non-users. She also said the company is structuring data so outside agents can read every authenticated, attributed, priced item. That is the same bet Vestiaire is making with OpenAI, only through Google’s door.

The quarter underneath that pitch was large enough to matter. The RealReal reported third-quarter GMV of $642 million, up 23% from a year earlier, and revenue of $206 million, up 19%. Orders rose 15%, average order value was $629, and trailing-twelve-month active buyers reached 1,152,000. Keyword search, Levesque’s team keeps saying, was built for retailers who sell many units of the same SKU. Resale does not work that way. Once a dress sells, it is gone.

Brands Now Compete With Bags They Already Sold

First-hand houses lose more than a single checkout when an agent can fetch last season’s bag as easily as this season’s. They lose control of the photo, the price next to it, and the seller standing behind it. They also hand resale platforms a demand signal their own till never captured.

Biddle said resale tells brands what people still want years after a piece left the shop floor, and that a spike in searches for an era or a label is often an early sign it is coming back. Friedland said Poshmark’s models already watch searches, listings, likes and sales by designer, era, silhouette and price, then set those against scarcity and actual resale prices to separate a fad from a piece that holds.

Bouchez still talks about trust as Vestiaire’s advantage, the thing it wants to hold while AI cuts the cost of checking. The platforms can polish a listing, write its caption, negotiate its offer and walk it up to ChatGPT. The agent then chooses three. The rest of the closet, however carefully tagged, does not get a vote.

Harry is the editor and lead writer of KERALANEWS 24X7, which he owns and runs as an independent publication. After ten years in journalism as a reporter and then an editor, he treats a story as something that keeps its history rather than a page that is silently replaced. When a report is updated, the new material is added with the time it arrived, and earlier text that turned out to be wrong is corrected in the open under the site's public corrections policy rather than deleted. Readers in any time zone can see how a story developed. Publishing around the clock never shortens the checking: the primary filing, statement, transcript or dataset is located first, and every number is confirmed against it before it appears. The site covers news, business and technology, science and sports, and entertainment, lifestyle and travel, with auto and gaming reported to the same standard, all for an international readership. Reader mail goes to Harry rather than to a form, at support@keralanews247.com.

Click to comment

Leave a Reply

Your email address will not be published. Required fields are marked *

Trending