Get Your Products Into ChatGPT Shopping: A Merchant Guide

Your customer asks ChatGPT for a sustainable water bottle that fits a school backpack. ChatGPT answers with a handful of product cards — photo, price, ratings, link. Your product isn't among them. Not in tenth place. Not at all.

To get your products shown in ChatGPT, you need three things: a robots.txt that doesn't lock out OpenAI's search crawler OAI-SearchBot, product data the machine can read — ideally as a direct product feed following OpenAI's specification — and a shop system that keeps that data current. This guide walks through all three steps, field by field, for Shopify, WooCommerce and Shopware.

How ChatGPT picks products for shoppers

ChatGPT answers buying questions with product cards showing price, image and reviews. The selection is not advertising: OpenAI itself states that product results are "organic and unsponsored, ranked purely on relevance". You can't buy a slot — what counts is data that is good enough.

Where does that data come from? According to OpenAI's own help documentation, from three sources: structured metadata from first- and third-party providers, other third-party content — and the crawl of your website. By now there is a fourth, more direct source: the product feed merchants submit to OpenAI themselves. If you only get crawled passively, you leave it to the machine to work out what your products are. If you supply the feed, you decide.

Step 1: Let OAI-SearchBot in — configure robots.txt correctly

Before any feed helps, the crawler has to get through the door. OpenAI operates three different bots, and they are constantly confused with one another:

• OAI-SearchBot — the only one that matters for your visibility. It is what makes websites "surface in search results in ChatGPT's search features". Block it and you do not appear in ChatGPT search answers.
• GPTBot — collects training data for the models. Blocking it is a legitimate decision about training use — it costs you no shopping visibility.
• ChatGPT-User — fetches pages when a user clicks your product inside a chat. It does not crawl automatically; per OpenAI, robots.txt rules may not even apply to it.

The correct allow rule looks like this:

User-agent: OAI-SearchBot
Allow: /

That sounds trivial. It isn't. We checked the copy-paste snippets of the top-ranking German-language guides on this topic against OpenAI's own crawler documentation: at least one of the top-3 results recommends a user-agent token that OpenAI does not define anywhere (checked 15 August 2026). Copy that snippet and you are configuring a rule for a bot that doesn't exist.

And the shops themselves? On 15 August 2026 we checked the robots.txt of 23 large German online shops — from department-store platforms to well-known DTC brands. The result: 19 of 23 do not mention a single AI crawler. Three explicitly allow at least one — one of them the search crawler only. One deliberately blocks the training crawler. The large majority has simply made no decision at all — neither for nor against. No strategy. No block. No plan.

Not mentioned technically means allowed. But it also means: nobody has checked whether the CDN's bot protection silently serves those same crawlers error pages. In our sample, one large German retailer answered requests carrying OpenAI crawler identification with a 404 — while browsers received the same page normally. The robots.txt said nothing about it.

Step 2: The OpenAI product feed — required fields, formats, freshness

The direct product feed is the most precise source you can hand OpenAI. The feed specification is public — and considerably stricter than most guides reproduce it (read 15 August 2026).

OpenAI accepts delimiter files only: UTF-8, tab-delimited as .txt or .tsv, or comma-delimited as .csv — each also gzip-compressed. No XML. Anyone selling you an XML feed in Google Shopping format as a ChatGPT feed has not read the specification.

These 14 fields are required for every product:

• item_id — your unique item ID
• title — the product name
• description — the product description
• url — the product page
• brand — the brand
• image_url — the main image (JPEG/PNG over HTTPS; image URLs behind logins are rejected)
• price — the price
• availability — stock status (in_stock, out_of_stock, pre_order, backorder)
• is_eligible_search — may the product appear in search?
• is_eligible_checkout — is it enabled for checkout?
• seller_name and seller_url — who sells, and where
• target_countries and store_country — target markets and the shop's home country

On top come conditional requirements: gtin or mpn (unless you explicitly declare no identifier exists), availability_date for pre-orders, privacy policy and terms once checkout is enabled. A blanket GTIN requirement, as circulated in several guides, does not exist in the specification — nor does an 800×800-pixel image rule.

Then there are the fields that decide recommendations despite being optional: review_count, star_rating, store_review_count, store_star_rating, reviews, q_and_a, and shipping with structured transit times. OpenAI's developer guide says it plainly: recommended attributes like rich media, reviews and performance signals "improve ranking, relevance, and user trust". A feed that fills only the required fields is in the race. A feed with proof fields wins it.

On freshness: per the onboarding guide, OpenAI expects an initial sample feed for validation, then daily snapshots — with intraday price and stock updates available via API. A feed whose prices don't match the product pages damages exactly the trust signal it is meant to build.

Step 3: Submitting the feed — the application at chatgpt.com/merchants

The direct feed is not self-service. OpenAI currently grants access to "approved partners": you fill out the merchant form at chatgpt.com/merchants, OpenAI reviews, then your sample feed is validated, then daily snapshots begin. Whether and how quickly merchants from the DACH region are being approved is not publicly documented — the form itself asks for no region. Applying costs nothing and puts you into the review process; waiting costs the time your competitors are using.

Shopify, WooCommerce or Shopware: the path to a ChatGPT feed per platform

Whatever the system: step 1 — robots.txt — is not something any plugin does for you.

Shopify: automatic, but only with US reach

Shopify catalogs are integrated into ChatGPT directly. OpenAI states that for Shopify merchants "no additional work is required". The catch is in Shopify's own documentation: your store must sell to customers in the United States for the ChatGPT channel to apply — where the company sits doesn't matter, where it sells does. A shop delivering exclusively to Germany and Austria is not covered by the automation. For it, the same paths apply as for everyone else: crawler access plus the direct-feed application. If you do sell into the US, the channel is on by default and purchases run through your own Shopify checkout.

WooCommerce: no native path, but a mature plugin ecosystem

WooCommerce ships no OpenAI feed of its own. Established feed plugins have caught up: Product Feed Pro by AdTribes generates an OpenAI channel mapped to the specification fields including the eligibility flags; alternatives include WebToffee, Webkul and CTX Feed. Important: the plugin only produces the file. The application through the merchant form remains your step.

Shopware: exactly one dedicated extension — otherwise manual work

For Shopware 6 the store currently lists a single dedicated solution: the "OpenAI / ChatGPT Merchant Feed Generator" by GOLLE IT (Shopware 6.6.0–6.7.13, €12.99/month, feed regeneration every 15 minutes, last updated August 2026). In fairness: the extension is young — single-digit downloads, no reviews yet. The alternative is a manually maintained export profile mapping your product data onto the 14 required fields. Doable, but maintenance-heavy — and exactly where price and stock drift creeps in when nobody is watching.

Is ChatGPT Shopping even available in Germany?

The short answer: product display yes, in-chat purchase no.

Product cards have long been appearing in German ChatGPT sessions — try it with a buying question of your own. OpenAI publishes no official country list. Buying directly in the chat — "Instant Checkout" — launched in September 2025 explicitly for US users and US sellers. It was never announced for the EU. And in March 2026 OpenAI visibly changed course: the company is focusing on product discovery and letting merchants sell in their own checkouts again — Instant Checkout "did not offer the level of flexibility that we aspire to provide".

For a merchant in the DACH region this means: the purchase happens in your shop today, as it always has. What has moved is the stage before it — who makes the shortlist when the customer asks. That stage is now configurable, through crawler access and the feed. There is no deadline and no penalty for being late — only taken seats on a short list.

Optimizing for ChatGPT product search: the factors OpenAI itself names

If you want to optimize for ChatGPT's product search, start with the criteria OpenAI documents — not with guesses. The help documentation names, for merchant selection: availability, price, quality, and whether the seller is the maker or primary seller of the item. The feed documentation adds reviews, rich media and performance signals as ranking factors.

That yields an uncomfortably simple priority list:

1. Reachability before optimization. A blocked crawler makes every further measure pointless.
2. Data quality before copywriting. Complete required fields, consistent prices, honest stock levels.
3. Proof fields before ad copy. Review counts and shipping data are machine-readable trust arguments — adjectives are not.
4. Consistency between feed and product page. The machine reads both and compares.

Beyond ChatGPT: Google AI Mode, Perplexity, Copilot

ChatGPT is the most visible AI surface recommending products, but not the only one. Google feeds its AI Mode from the Shopping Graph — maintained through Google Merchant Center. Perplexity runs its own merchant program. Microsoft's Copilot draws on the Bing index and Microsoft Merchant Center. The discipline is the same everywhere — clean structured data, reachable crawlers, current feeds. Do the homework for ChatGPT and the largest part of the work for the other three surfaces is already done. Each program deserves its own guide; the order above is the shared starting point.

Frequently asked merchant questions

Does it cost anything to appear with products in ChatGPT?

No. Product results are organic and unpaid per OpenAI; there is no paid placement. A fee applies only to purchases through Instant Checkout — which currently exists only in the US.

Do I have to submit a feed, or does ChatGPT find my shop on its own?

By default ChatGPT works with crawled data and third-party metadata. That works — but you don't control it. The direct feed gives you control over titles, prices, availability and proof fields, and is the most direct way to guarantee freshness.

Which fields must the OpenAI product feed contain?

14 required fields per product: identity (item_id, title, description, brand), page and image (url, image_url), price and stock (price, availability), the search and checkout eligibility flags, seller (seller_name, seller_url) and target markets (target_countries, store_country). GTIN or MPN are only conditionally required; an 800×800 image rule does not exist in the specification. The annotated list is in step 2 above.

Which crawler do I need to allow in robots.txt?

OAI-SearchBot — it alone controls visibility in ChatGPT search. GPTBot affects model training only; ChatGPT-User covers user-initiated page visits.

How do I know whether ChatGPT sends me visitors?

Visits from ChatGPT appear in your analytics as referral traffic from chatgpt.com. Setting up a dedicated segment is the fastest way to quantify the channel from day one — before it is big enough for someone to doubt it.

Does this work if I only sell to Germany?

Yes — via crawler access and the direct feed. Only the Shopify automation requires US customers. target_countries in the feed declares your markets explicitly.

Getting your products into ChatGPT: three steps, no magic

Getting products shown in ChatGPT is not magic and not a lottery. It is data infrastructure: crawler in, feed out per specification, shop system connected, proof fields filled. And the finding from our sample — 19 of 23 large German shops without any AI-crawler decision at all — also means: the starting line is still remarkably empty.

If you want to know where you stand today: with our free AI Visibility Check we check what ChatGPT, Google AI Mode, Perplexity and Copilot recommend for three realistic buyer prompts in your market — against a competitor you name, with a short video and a one-page summary. No call required.

Further reading:

• GDPR-compliant AI for German companies — why data architecture is half the battle here too
• Private AI vs. cloud AI — where your data sits when machines read it
• Our AI visibility offer at a glance

---

All OpenAI and Shopify statements in this article were checked against the linked primary sources on 15 August 2026. The platforms change their programs continuously — verify the linked specifications before implementing.