Getting Your Products Into ChatGPT Shopping Isn't Enough

Getting your products into ChatGPT shopping takes two moves. Apply at chatgpt.com/merchants, or do nothing at all if you already sell through Shopify or Etsy, both of which are auto enrolled. Submit a feed OpenAI can refresh as often as every 15 minutes. That makes you eligible. It does not make you the one ChatGPT recommends.

How to get your products into ChatGPT shopping

OpenAI runs enrollment through the Merchant Program at chatgpt.com/merchants. Shopify and Etsy sellers skip that step entirely, both platforms wire their full catalogs into ChatGPT automatically, no separate application required. What still decides whether a product actually shows up is the feed itself: submit data in CSV, TSV, XML or JSON, keep price, availability and image fields current, and OpenAI pulls refreshed data as often as every 15 minutes. Shopping inside ChatGPT is currently live for United States users, and OpenAI has said it is building a self serve platform to widen access to more merchants and regions over time.

That checklist is the easy half. A complete feed makes a product eligible to be shown. It says nothing about whether ChatGPT chooses to recommend that product over the next nine sellers running the same feed.

Why a feed does not get your product chosen

Azoma analyzed millions of ChatGPT shopping agent citations from Q2 2026 and published the source breakdown in late August. Earned media, independent reviews, roundups and articles the brand does not control, accounted for 41 percent of citations. Retailer listings accounted for 37 percent. User generated content accounted for 19 percent. A brand's own site, the page a seller controls completely and usually fills out the most carefully, accounted for 3 percent.

We run a version of the same check on our own domain, though our tracker currently watches Claude, not ChatGPT, so treat this as a related signal rather than a repeat of Azoma's number. Between July 2 and September 8, 2026, we ran 1,257 prompt checks asking named questions about specific products, drops and objects across fashion, music and culture. Finallyoffline.com turned up as a cited source in 21 of those checks, about 1.7 percent. In every one of those 21 answers, our page sat in a source list next to a mix of brand owned pages and independent press, outlets like Hypebeast, Complex, Culted and GuitarWorld. The brand's own domain was rarely the only source in the answer and never the exclusive one. That is consistent with what Azoma found for ChatGPT: a feed or a brand page fills one slot. The other slots go to whoever already wrote about the product.

If your product only exists on your own store page, you are competing for 3 percent of the citation pool with everyone else who also only has a store page. The other 97 percent belongs to coverage nobody at your company wrote.

What actually earns the other slots

The causes are the same ones that keep any page out of an AI answer, they just show up faster in shopping because the competition for each slot is denser. A single product page with a description and a price is not the same as an article a customer can quote back. Answer engines lift structured, complete pages, ones that name the product, answer real questions about it, and read like something a person wrote to be understood, not indexed. We found the same pattern outside shopping: AI answers pull from websites almost exclusively, and the pages that get pulled are structured articles, not feeds or social posts.

The other cause is volume. One well written page about a product is a lucky citation. A domain that consistently publishes structured coverage, its own drops, its own comparisons, its own answers to the questions buyers actually type, gives an answer engine more chances to pick it up. Our own archive ranks and gets clicked in Google's results for exactly this reason, and the same structure is what AI shopping agents are scanning for when they fill that 41 percent earned media slot.

Turning what a five person brand already has, its posts, its drops, its founder notes, into that kind of page is the actual fix, and it is the part a product feed cannot do for you. If your product updates already live on Instagram, turn those posts into the structured articles ChatGPT and Google both pull from.

What an earned media layer looks like for a five person brand

A five person brand does not need a newsroom to compete for that 41 percent. It needs a small, steady stream of structured pages that a Merchant Program feed cannot produce on its own: one article per new drop or restock explaining what changed and why, one page answering the three questions customers actually ask before buying, and a running archive so an answer engine has more than one page to choose from. We built our own version of this pipeline by turning Instagram captions into ranked articles, which is the same shape of work a small commerce team needs, just pointed at products instead of culture coverage. The feed gets a product listed. The archive is what gets it chosen. Turn your product updates into the structured articles that earn those citations.