# ChatGPT Shopping: How to Get Your Products Recommended

> ChatGPT Shopping leans on product feeds. Fix feed eligibility first, then the product facts ChatGPT repeats, then reviews. A checklist and tracking plan.

- Source: https://www.citlyze.com/blog/chatgpt-shopping-product-recommendations
- Author: Mara Ellison
- Published: 2026-10-02
- Tags: Strategy, Engines

To get products recommended in ChatGPT Shopping, get into its product data first. Shopify and Etsy catalogs are already connected; other merchants apply for a direct feed. Then make the facts ChatGPT repeats (price, stock, specs, ratings) complete and consistent, and work on reviews last. On shopping prompts, the feed now outweighs your content.

That order runs against the usual GEO advice, which starts with content. The reason is a change in where ChatGPT gets its product picks, which OpenAI's own help pages point toward and which trade press reported in September from one vendor's tracking data.

## How ChatGPT chooses which products to show

ChatGPT picks products from structured product data plus third-party content, weighted by what the shopper asked for. Since July, most picks appear to come from feed-integrated sources.

OpenAI's [help page on shopping](https://help.openai.com/en/articles/11128490-shopping-with-chatgpt-search) lists what ChatGPT considers:

- "Structured metadata from first-party and third-party providers (e.g., price, product description) and other third-party content."
- The model's own response before any new search.
- OpenAI's safety standards and product policies.

For a given question it may weigh "available options, price, reviews, and ease of use", plus the user's memory and custom instructions. It can rewrite product titles, add labels such as "Budget-friendly" based on what reviewers say, and summarize reviews from public websites. When several merchants sell the same item, it ranks them by "availability, price, quality, and whether they are the maker or primary seller". OpenAI states that product results "are not ads, nor influenced by any OpenAI partnerships." Advertising on ChatGPT does not change your shopping placement either; ads appear separately below answers, as [the guide to ChatGPT ads](https://www.citlyze.com/blog/chatgpt-ads-sponsored-agents-ai-mode-ads) explains.

The tracking data comes from an AI visibility vendor, Profound, whose September analysis [Search Engine Journal](https://www.searchenginejournal.com/chatgpt-shopping-results-lean-hard-on-product-feeds/589000/) reported on September 9. Across 1,757,723 tracked prompt runs in July, feed-integrated product picks rose from 8.26% to 61.54% of recommendations on July 10, 2026, and reached about 65% by early September. Around the same date, the top 10 merchants' share of references rose from 22.5% to 41.8%, and the number of distinct merchants referenced fell from 13,524 to 10,607. Of 687 tracked customers, 450 saw their Shopping visibility fall by a third or more between the three days before and the three days after July 10, against 67 that rose by that much, and the analysis estimated that Shopify accounted for about 35% of feed-integrated retrieval.

Read those numbers with their limits. They come from one vendor's tracked prompts, not from real shopper sessions, and we have not reproduced them. OpenAI has not confirmed a change on July 10, and Search Engine Journal noted that ChatGPT's release notes list no shopping update for July 9 or 10.

| Input ChatGPT uses | What OpenAI says | What you control |
| --- | --- | --- |
| Structured product data | Price, description and other metadata from first-party and third-party providers | Your feed, or your Shopify or Etsy catalog |
| The shopper's context | Query, memory and custom instructions | Nothing directly; cover the conditions buyers state, such as budget or size |
| Reviews | Summaries from public websites; unverified by OpenAI | Review volume and content on your site and on retailers |
| Merchant ranking | Availability, price, quality, maker or primary seller | Stock accuracy, competitive pricing, selling direct |

## Whether your catalog is eligible for ChatGPT's product feed

If you sell through Shopify or Etsy, your catalog is already integrated. Everyone else applies through OpenAI's merchant page and joins a waitlist, and shopping itself is live only in the US for now; OpenAI says it will expand to more regions over time.

OpenAI's [merchant page](https://chatgpt.com/merchants/) spells out the terms: "If you sell through Shopify or Etsy, your catalog is already integrated, and no additional setup or application is required." Other merchants fill in an application; a self-serve merchant portal is planned for later this year. There are no fees on purchases that start in ChatGPT, and OpenAI says it is "moving away from a standalone Instant Checkout experience", so shoppers discover products in ChatGPT and complete the purchase on your own site or app.

The same page answers a common question: do you need a feed if ChatGPT already crawls your site? "No, but feeds give you greater control over how your products appear."

If you prepare a direct feed, OpenAI's [product feed specification](https://developers.openai.com/commerce/specs/file-upload/products) is the checklist. Work through it in this order:

1. Confirm your route: Shopify Catalog, Etsy, or a direct feed application. Shopify merchants can use the remaining steps to audit the product data their store already sends.
2. Fill the nine required fields for every item or variant: `item_id`, `title`, `description`, `url`, `brand`, `seller_name`, `image_url`, `availability` and `price`.
3. Write prices as an amount and a currency code, such as `79.99 USD`, and update them when a sale starts or ends. Use `sale_price` only when it is lower than `price`.
4. Set `availability` to one of the listed values. OpenAI says an omitted, empty or unrecognized value rejects the row; use `unknown` explicitly when you do not know.
5. Model variants with a stable `group_id` shared by the variants, `listing_has_variations=true`, and a `variant_dict` such as `{"color":"Black","size":"10"}`.
6. Add identifiers you hold: a valid GTIN with its check digit, or an MPN together with the brand. The spec warns against inventing an MPN to cover a missing GTIN.
7. Add product-level reviews as `review_count` and `star_rating` from the same review population. Store reviews go in separate fields.
8. Fill shipping and returns (`shipping_price`, `accepts_returns`, `return_deadline_in_days`, `return_policy`), and keep `is_eligible_search` true.
9. Keep product and image URLs public and stable, with the selected variant in the URL where possible.

The spec runs to 48 fields in total and accepts Google-compatible feeds, so a clean Merchant Center feed gives you a head start.

## The product facts ChatGPT repeats

ChatGPT repeats the facts it can read in structured data and on public pages: price, availability, title, key specs, ratings and the themes in reviews. Keep each one identical everywhere it appears.

| Fact | Where ChatGPT reads it | How to keep it consistent |
| --- | --- | --- |
| Price | Feed or third-party providers; OpenAI notes updates can lag | Update the feed when prices change, and match the product page |
| Availability | Feed `availability` field | Sync stock at least daily, and mark sold-out variants `out_of_stock` |
| Title | Your title, often simplified by ChatGPT | Put brand, product type and the defining attribute first |
| Specs and attributes | Feed fields such as material, dimensions, color | Use the structured fields, and state the same values on the page |
| Ratings | `review_count` and `star_rating`, plus public review sites | Keep one review population per product; show it on the page |
| Labels such as "Budget-friendly" | Generated from reviews and third-party data | Answer value and quality questions in your own copy |

Product pages are the weak link for many retailers. Adobe's April 2026 analysis found U.S. retail [product detail pages scored 66%](https://business.adobe.com/blog/ai-traffic-surge-retail-sites-not-machine-readable) on its AI readability measure, meaning about a third of their content was not readable by LLMs, below homepages at 75%. Facts that sit in images, tabs rendered by script, or PDFs are facts ChatGPT may never see.

## The third-party signals that matter after the feed

Reviews and the retailers that stock you matter most, because ChatGPT summarizes public reviews and ranks merchants per product.

- Reviews on your site and on retailer pages feed the summaries and labels ChatGPT writes. OpenAI states it does not verify them, so volume and recency on the pages it can read carry the weight.
- Retail and marketplace listings. Similarweb reported that [54.7% of sources cited in ChatGPT beauty conversations](https://aisearch.similarweb.com/blog/ai-referral-traffic-by-industry/) were retail and e-commerce domains (US desktop, May 2026), and that AI answers favor large multi-category platforms. If you sell through marketplaces, your listings there are part of your AI visibility.
- Editorial reviews and buying guides. These shape the model's answer before it looks at products, which is the second input on OpenAI's list. Our explainer on [ChatGPT's source selection](https://www.citlyze.com/blog/how-chatgpt-picks-sources) covers that layer.
- Accuracy across all of them. A wrong price or a discontinued spec on a retailer page can surface as a fact about your product. [Negative or wrong brand claims in AI](https://www.citlyze.com/blog/fix-negative-brand-sentiment-ai) walks through fixing the source.

## Tracking ChatGPT Shopping visibility by category

Track it on the consumer app, per category, with a fixed prompt set, because product carousels appear only in consumer ChatGPT and vary between runs.

- Build a prompt set per category: generic ("best trail running shoes under $150"), attribute-led ("waterproof trail shoes wide fit"), and comparison prompts naming your products.
- Record per run: whether shopping results appeared at all, which SKUs showed, their position, the merchant and the price shown.
- Compute visibility rate per SKU (runs showing it divided by runs observed) and a shopping trigger rate per prompt. Read them over several weeks; [AI answers change between runs](https://www.citlyze.com/blog/why-ai-answers-change), so one check is not a measurement.
- Log feed changes with dates next to the results, so you can see whether a fixed field or a price change moved anything.
- Compare categories. The July data showed visibility concentrating in fewer merchants, so a category where you hold steady may hide one where you have dropped out.

To see which of your SKUs ChatGPT and Google's AI surfaces show for your buying prompts, Citlyze's [AI Shopping](https://www.citlyze.com/features/ai-shopping) matches your catalog (Shopify, a Merchant Center feed or a CSV) against tracked answers and scores each product's visibility, position and win rate, plus the merchants and prices shown next to it. For the brand-level view outside shopping carousels, [tracking brand mentions in ChatGPT](https://www.citlyze.com/blog/track-brand-mentions-chatgpt) covers mention and citation tracking.

## Fix the feed, then the facts, then the reviews

Check that your catalog reaches ChatGPT, that each product's price and stock match everywhere, and only then invest in reviews and content. Citlyze's [AI Shopping](https://www.citlyze.com/features/ai-shopping) tracks up to 100 SKUs on the Growth plan across Google AI Mode and AI Overviews, with ChatGPT carousel capture as an add-on, and every plan shows where your brand stands in ChatGPT's regular answers. Plans start at $29 a month with a [14-day trial and no card](https://www.citlyze.com/pricing).

---

Canonical: https://www.citlyze.com/blog/chatgpt-shopping-product-recommendations