Buyer Agents Are Coming. Every Fashion Brand Needs a Seller Agent

Shoppers are starting to hand their shopping to AI. They ask ChatGPT, Gemini or Perplexity to find a linen shirt under $80 or a wedding-guest outfit that ships to Dubai, and the assistant searches, compares, shortlists and, more and more often, buys. Call these assistants buyer agents: software that shops for a person.

On the other side of every one of those purchases sits a brand, and today a brand meets a buyer agent with a product page and a data feed. That is enough for a phone charger. It is not enough for a kurta. Fashion orders turn on questions a feed can't answer: does it run small, is the fabric sheer, can it be stitched to my measurements by the 20th, can I exchange the size. A brand needs something that answers for it, from its own catalogue and policies, every time a buyer agent asks. That is a seller agent, and our view is that every fashion brand will need one.

This piece sets out the shift, why fashion feels it first, what a seller agent does, and what it means for brands and for the teams building shopping agents. We make TailorTalk, which runs seller agents for fashion brands, so read it with that in mind.

What is a buyer agent, and what is a seller agent?

A buyer agent is an AI assistant that shops on a person's behalf. The shopper says what they want, and the agent searches across stores, compares options, answers the shopper's questions and, where it is allowed to, places the order. ChatGPT, Google's AI Mode, Gemini, Perplexity, Microsoft Copilot, Amazon's Alexa for Shopping (formerly Rufus) and Meta's new Muse all do parts of this today.

A seller agent is the brand's side of that conversation. When a buyer agent needs to know something the product page doesn't say, the seller agent answers from the brand's own catalogue, live stock and policies. It only speaks for that brand, never recommends a competitor, and shares only what a shopper could see on a product page.

Buyer agents are already here

In the last year every large AI assistant has started to shop. Here is what each one does today, as of October 2026:

The money is following. Adobe measured traffic from AI assistants to US retail sites up 693.4% over the 2025 holiday season, and by July 2026 those visitors converted 60% better than other retail traffic. Salesforce estimates AI and agents drove $262 billion of global holiday sales in 2025. Morgan Stanley found nearly 23% of Americans had made at least one AI-assisted purchase in the past month, and projects agentic shopping could drive up to $385 billion of US online sales by 2030; McKinsey puts the US figure at up to $1 trillion of orchestrated revenue by 2030.

Two things stand out. First, the assistants are converging on the same shape: the agent finds and compares, the brand stays the merchant and owns the checkout. OpenAI went furthest towards buying inside the chat and then stepped back. Second, the fights are about access. Amazon blocked Meta's Muse from Amazon.com within two weeks of its launch, and it has been in court with Perplexity over its shopping agent since November 2025. Brands that welcome agents, and give them good answers, will be the easy ones to buy from.

Why fashion feels it first

Most product categories fit in a feed. A charger has a wattage, a price and a delivery date, and a buyer agent can compare a hundred of them without asking anyone anything. Clothes don't work like that:

• Sizes don't travel. An M at one label is an L at another, and "runs slim" lives in the head of the person who packs the orders, not in the feed.
• Fabric and drape decide the sale. "Is the chanderi sheer?", "Does it come with a lining?", "Will the colour bleed?" are answered by the brand or not at all.
• Much of the catalogue is made, altered or stitched to order. Ethnic wear in particular sells as stitched, semi-stitched or unstitched, with custom sizing and a lead time.
• Stock is by size and colour, and it moves daily. "Is the rust one still there in L?" has a different answer every afternoon.
• Delivery dates matter more than in most categories, because outfits are bought for an event.

The cost of getting this wrong is already high. Coresight Research estimates the average US online apparel return rate at 23.4% in 2025, and in its survey nearly 70% of shoppers who returned clothing bought online gave size and fit as the reason (a study sponsored by the sizing company Alvanon). A buyer agent that guesses the size makes that worse; one that can ask the brand makes it better.

These are the questions that already fill fashion brands' Instagram DMs and WhatsApp chats. A buyer agent will ask the same ones, on the shopper's behalf, and it will move on to the next brand if nobody answers.

Why a product feed is not enough

Feeds and product pages are built to answer "what is it and what does it cost". They are one-way: the brand publishes, the agent reads, and anything not written down is a guess. When a buyer agent hits a question the feed can't answer, it has three bad options: guess (and risk a return), drop the brand from the shortlist, or send the shopper off to read the brand's pages themselves, which is the work they asked the agent to do.

A seller agent turns the feed into a conversation. The buyer agent asks; the brand answers, correctly and at once, from data that is current to the minute. For the shopper, that means fewer wrong sizes. For the brand, it means staying on the shortlist, and keeping its voice and its rules in a sale it no longer sees directly.

What a seller agent does

• Answers from the brand's own catalogue and policies only: fit, fabric, stitching, stock by size, delivery dates, cash on delivery, exchanges and returns.
• Speaks only for that brand. If a piece is out of stock, it suggests the closest match from the brand's own range, never another label.
• Shares what a product page would show, and nothing else. Customer conversations, contact details, order history and sales numbers stay private.
• Works the same everywhere. The agent that answers a buyer agent is the one already answering shoppers on Instagram, WhatsApp and the brand's website, so every channel gets the same answer.
• Knows when to hand over. Custom orders, complaints and big wholesale enquiries go to a person on the team.

What this means for fashion brands

Brands don't need to understand any of the plumbing. They need three things: a catalogue an AI can read, prices, stock and policies that are always current, and an agent that answers for them. The practical side, from writing product pages the way shoppers ask to checking whether ChatGPT recommends you, is in our guide to getting your fashion brand recommended by ChatGPT.

TailorTalk's AI shopping assistants feature, now in early access, does the rest: it makes a brand's catalogue, sizes, live stock and policies readable to assistants such as ChatGPT, Gemini and Claude, and its TailorTalk agent answers their questions on fit, stitching, delivery and exchanges. It works for brands on Shopify, WooCommerce or Magento, and for boutiques with no website at all, through a catalogue built from photos with AI Catalog.

What this means for teams building shopping agents

If you build a buyer agent, fashion is where your users will judge you, and it is where generic product data lets you down. Three problems come up again and again:

• Coverage. A large share of fashion, especially independent and ethnic wear labels, sells mainly through Instagram and WhatsApp, with no store feed to read. Those brands are invisible to an agent that only reads feeds.
• Depth. Feeds carry titles, prices and a size list. They rarely carry fabric weight, fit notes, garment measurements, stitching options or exchange rules, which are what a shopper actually asks.
• Freshness and answers. Stock by size changes daily, and many questions have no field at all. An agent needs a way to ask the brand and get a reliable answer back.

Platforms are starting to fill part of the gap: Shopify, for one, now hands its merchants' catalogues to ChatGPT, Copilot and Google by default. That covers the brands already on Shopify, and the data their feeds already hold. It doesn't cover the boutique that sells on Instagram, or the question no field answers.

A seller agent on the brand's side solves all three: structured fashion data, live stock, and an answer to any question the data doesn't cover, in one place for every brand on it. We are opening TailorTalk's brands to shopping agents on that basis. If you are building one, get in touch through the AI shopping assistants page.

Why we are building this

TailorTalk already is a seller agent, for shoppers rather than for AI. Fashion brands use it to answer Instagram DMs, WhatsApp messages and website chats around the clock, from their own catalogue, with live sizes and stock. Samyakk, an ethnic wear brand, handles more than 1,000 Instagram inquiries a day with it; Mysore Saree Udyog handles more than 8,500 leads a month across WhatsApp and Instagram. The questions buyer agents will ask are the ones these agents answer every day.

Pointing the same agent at AI assistants is the next step: one source of truth about a brand, answering whoever is asking, a shopper in the DMs or an assistant shopping for one. For how shoppers already buy in chat, see our guide to conversational commerce.

References

• OpenAI: Powering product discovery in ChatGPT (March 2026), and ChatGPT merchants FAQ.
• Google: Agentic checkout and AI shopping tools (November 2025); new tools and an open protocol for agentic commerce (January 2026).
• Perplexity: Shopping that puts you first (November 2025).
• Engadget: Microsoft is integrating shopping directly into Copilot (January 2026).
• Amazon: Rufus, now Alexa for Shopping, and Buy for Me.
• Meta: Introducing Muse (September 2026).
• Adobe: 2025 holiday shopping season results (January 2026), and AI Traffic Trends Report (August 2026).
• Salesforce: 2025 holiday shopping data (January 2026).
• Digital Commerce 360 on Morgan Stanley Research: AI agentic shoppers and $385 billion in US online sales (December 2025).
• Shopify: agentic commerce momentum (March 2026).
• FashionUnited (Alvanon) on Coresight Research, Shifting the Size and Fit Paradigm (2026).
• McKinsey: The agentic commerce opportunity (October 2025).