How to Sell on ChatGPT: A Brand's Guide

ChatGPT is a shopping surface now

Somewhere along the way, "ask ChatGPT" quietly replaced "Google it" for a lot of buying decisions. People ask it for gift ideas, skincare routines, running shoes for bad knees, and laptops under a budget. AI assistants handle over 5 billion queries a day across ChatGPT, Google AI, and Perplexity, and a growing slice of those are purchase decisions in progress.

Then OpenAI made it official. The Agentic Commerce Protocol (ACP) gives ChatGPT real commerce rails: product feeds, carts, orders, and authentication. A shopper can go from "what should I buy" to "order placed" without ever opening a browser tab. Adobe has tracked AI-driven retail traffic growing about 4,700% year over year, so the buyers are already there.

The question is whether your products are.

How ChatGPT picks products, in plain terms

It helps to drop the mental model of a search results page. ChatGPT is not ranking ten blue links. It is composing an answer, and it only includes products it can stand behind.

When someone types "cute tops for a brunch date, something flowy and flattering under 1k," three things happen fast:

• The question fans out. One request becomes many: flowy tops under 1,000, brunch outfit tops, flattering casual tops, best rated tops this season. Each sub-query goes looking for evidence.
• The evidence is your data. Those sub-queries match against structured product feeds. Category, fabric, fit, occasion, price, stock, reviews. The model does not infer what your product might be. It reads what your data says it is.
• The winners become the answer. A handful of products come back as cards in the chat, with prices and a path to buy. Everything else, including products that would have been perfect, simply never appears.

We went deeper on the failure side of this in why ChatGPT doesn't recommend your products. The short version: thin data is invisible data.

What the buying experience looks like

It is worth seeing the whole flow from the shopper's side, because it explains why this channel converts differently.

The shopper asks. ChatGPT replies with product cards: image, price, a line on why each fits. The shopper pushes back the way they would with a good salesperson. "Does it run small?" "Is this safe for sensitive skin?" "What colours does it come in?" The AI answers from enriched product data and reviews pooled from marketplaces and communities. Doubts get handled in the moment they appear, instead of becoming abandoned carts.

Then the part that changes the economics: the shopper says "add this to my cart" and checks out right there. Address, delivery, payment, confirmation, all inside the conversation. No redirect to your website, which means no drop-off between decision and purchase. That single step is what agentic commerce is really about.

What your brand actually needs

Strip away the buzzwords and selling on ChatGPT comes down to three requirements.

1. A feed built for an AI, not a search box

Most product feeds carry five to eight fields: title, price, image, link, maybe a category. An AI assistant comparing products wants 15 to 40 per SKU, depending on category. Fabric and fit for apparel. Ingredients and skin type for beauty. Battery life and compatibility for electronics. Plus the context that makes attributes mean something: who this is for, which occasion, which problem it solves. That work is the core of product feed optimization for AI search.

2. Transaction endpoints

Discovery without checkout is a brochure. To sell inside the chat you need the commerce plumbing ACP expects: cart building, offers, identity, and order handling that work in-conversation. This is the piece most brands cannot build quickly in-house.

3. Trust signals

ChatGPT puts its own credibility on the line with every recommendation, so it leans on proof: ratings, reviews, FAQs, user content, repeat-purchase signals. A product with rich data but no visible trust often loses to a slightly worse product with strong proof. Aggregating reviews from the marketplaces and communities where they already live is low-hanging fruit.

The fastest path: connect once

You could build all of this yourself: enrich every SKU by hand, stand up ACP endpoints, aggregate reviews, then keep up as the spec changes. Some teams will. Most should not, because the protocols and required fields shift week to week, and ChatGPT is only one of the surfaces that matter. Gemini, Perplexity and Google AI Mode, WhatsApp, and independent AI agents all want the same structured data in slightly different shapes.

That is the gap Ziffi exists to close. It pulls your existing catalog, enriches it into an intent-native product graph, syncs it to every AI surface including ACP, and handles native checkout. Your store and your customer experience stay untouched, and you keep full control over which platforms carry your products. Most merchants are live within hours. Early movement tends to show within a couple of weeks, with the measurable impact arriving in the 60 to 90 day window.

A 30-day starting plan

1. Week 1: measure. Ask ChatGPT the ten questions your best customers would ask. Note which brands come back and whether you are one of them. For a structured version of this, see AI visibility and share of voice.
2. Week 2: audit the data. Take your top 20 SKUs and count real attributes per product. Under ten means you are functionally invisible to comparison queries.
3. Weeks 3 and 4: close the gap. Enrich the catalog, connect the feed, switch on in-chat checkout. One integration if you use infrastructure, a roadmap if you build.

The brands that treated Google SEO seriously in 2005 owned a decade of cheap demand. The same window is open right now on ChatGPT, and it will not stay open long.