How an online store gets into ChatGPT answers in 2026

None of them is a switch you flip: each platform decides what it shows, and says so in its own documentation. What you can work on is whether an assistant can read your store, whether what it reads is complete, and whether other sites vouch for you. Below are the checks for each from our launch checklist, and what changes when Shopify’s channel leaves your products out. The platform facts were read on 23 October 2026.

The short answer

An assistant reaches products by two roads. One is a product channel: on Shopify, Shopify Catalog sends your product data to ChatGPT and Microsoft Copilot, with nothing to set up if your store is eligible. The other is the open web: assistants search for pages, read them and cite them, much as a search engine does. The work on the open web is the same on every platform:

1. Let the search crawlers in, in robots.txt and at the firewall.
2. Put the product facts in the HTML the store sends.
3. Answer shoppers’ questions as text on a page.
4. Be mentioned on sites you do not own.

Then measure what can be measured.

Two ways into an AI answer

The product channel: Shopify Catalog and Agentic Storefronts

Shopify’s agentic storefronts put your products in “AI channels, such as ChatGPT, Google AI Mode and Gemini, Microsoft Copilot, and Meta”, and the feature “is active by default for eligible stores” (Shopify Help). ChatGPT and Copilot read your products from Shopify Catalog; Google AI Mode and Gemini get them through the Google & YouTube sales channel or Google Merchant Center, and Meta through the Facebook and Instagram by Meta channel. You choose the channels in Sales channels > Agentic.

For ChatGPT, Shopify asks that the store sells to customers in the United States (it can be based anywhere), that the products are eligible for Catalog, that you accept Shopify’s agentic storefronts terms, and that your terms of service, privacy policy, and return and refund policy are filled in under Settings > Policies (Selling on ChatGPT). The shopper pays in your own checkout, opened inside ChatGPT or in a new tab. OpenAI says shopping is “currently live for ChatGPT users in the U.S.” (OpenAI), so for buyers in Canada, Australia or Europe this channel is not yet the road in.

The channel carries what your admin holds: each product’s “title, description, options, images, price, availability” and more, kept current (Shopify Help). To see where you stand, open Sales channels > Agentic: next to Sources > Shopify Catalog, Shopify shows how many of your products are in Catalog, or an Ineligible warning with the reason.

The open web: the pages assistants search and read

Shopify says your products are also discoverable through “other methods such as web crawling and indexing”, and that with Catalog turned off for ChatGPT they might still show there (same page). When ChatGPT picks products, OpenAI says it weighs structured metadata such as price and description “from first-party and third-party providers” and other third-party content, and its review summaries draw on reviews from public websites (Shopping with ChatGPT Search). To make a site eligible for ChatGPT search, OpenAI asks you to allow OAI-SearchBot and to let its published IP addresses through your host or CDN (Searching the web with ChatGPT). Google asks for nothing special: for AI Overviews and AI Mode, “there are no additional requirements” beyond being indexed and eligible to show with a snippet (Google Search Central). One setting can still keep a site out: Search Console’s Search generative AI control, under Settings, rolled out to all sites by August 31, 2026. It is on Include by default, and a site switched to Exclude does not appear in AI Overviews or AI Mode (Search Console Help).

This road matters for every store. For stores that Catalog leaves out, it is the only one.

Not on Shopify: product feeds and Shopify’s Agentic plan

For ChatGPT, a store on WooCommerce or another platform can send its product data two ways. OpenAI takes applications to share a product feed directly; applicants wait on a list, and OpenAI says it is building a self-serve platform for merchants (OpenAI). Shopify’s Agentic plan, public since March 2026, lets brands that do not use Shopify for ecommerce add their products to Shopify Catalog (Shopify). It asks for an external product URL on every product, registration documents and an identity check (Shopify Help).

Step 1: let the right crawlers in

Training, search and user-triggered crawlers are different

AI companies run several crawlers, each with its own job. OpenAI’s show the pattern (Overview of OpenAI Crawlers):

• OAI-SearchBot finds pages for ChatGPT search. A site that blocks it “will not be shown in ChatGPT search answers”, except as a bare navigational link.
• GPTBot collects pages for training. You can block it and still allow OAI-SearchBot.
• ChatGPT-User opens a page when a user’s question calls for it, and OpenAI says robots.txt rules “may not apply” to it.

Anthropic runs the same three kinds: ClaudeBot for training, Claude-SearchBot for search and Claude-User for user requests (Anthropic).

Our AI visibility check reads your robots.txt for each crawler in the table below, by the rules of RFC 9309, the robots.txt standard, and says for each whether it may read your home page:

Blocking a training crawler such as GPTBot or ClaudeBot is your call: it does not decide whether you show up in search answers, while blocking a search crawler such as OAI-SearchBot or PerplexityBot can keep your pages out of them. Google-Extended is different: Google says it also controls whether your pages are used to ground answers in the Gemini apps, though not whether they appear in Google Search (Google’s crawler list). Our checklist records the client’s written decision for each named AI crawler before anyone edits robots.txt, because otherwise someone blocks every AI user agent as a precaution (the checklist group).

Two rules from the standard: robots.txt is a request, not a lock, since RFC 9309 says its rules “are not a form of access authorization”, and a robots.txt that fails with a server error tells crawlers to assume “complete disallow” (RFC 9309). On Shopify, you change the rules by customizing the theme’s robots.txt.liquid template (Shopify Help), and it has to be on the live theme: rules edited on an unpublished copy never reach production.

A firewall can undo robots.txt

robots.txt can say yes while the server says no. A bot protection layer, a CDN rule or a “block bad bots” app can turn a crawler away before it reaches a page. Google’s basics for AI features include “ensuring that crawling is allowed in robots.txt, and by any CDN or hosting infrastructure” (Google Search Central).

So the checklist has a test before launch: request the home, collection and product templates on the live domain with the GPTBot, OAI-SearchBot, ClaudeBot and PerplexityBot user agents, and confirm each returns HTTP 200 with the full page. OAI-SearchBot is the one ChatGPT search depends on. A bot protection layer can answer everything except Googlebot with a 403, and the store is then invisible to AI crawlers. Borrowing a crawler’s name is a first test, not the whole one: OpenAI publishes the IP ranges of each of its crawlers and, for OAI-SearchBot, recommends allowing requests from them (Overview of OpenAI Crawlers).

On Shopify, network-level bot management is Shopify’s own, and Shopify does not recommend a proxy in front of the store, so look at any app or proxy added on top. Either way, blocking affects only the open web: Shopify notes it “doesn’t stop your product data from being sent by Shopify Catalog” (Shopify Help).

Step 2: put the product facts in the page

What a crawler gets with JavaScript off

Our check reads the HTML the store sends, the way a crawler does, so anything added later by JavaScript is not seen. The checklist’s version: load a product page with JavaScript off, and confirm the title, price, currency, stock state and full description are in the first response. Themes that fetch the price after load serve an empty shell, and the product reads as having no price.

You can test it without a tool: view a product page’s source (not the inspector, which shows the page after scripts ran) and search it for the price.

Specs as text, barcodes as GTINs, one Product block

Structured data (JSON-LD) is the markup search engines and AI assistants read for price, stock and the store’s name; our check looks for Product markup on a product page and Organization markup on the home page. The checks that matter most are in the checklist’s Catalogue and search pack:

• Specs as text. Materials, dimensions with units, weight, box contents, compatibility and care in metafields, shown as text on the product page. Specs inside an image or a PDF are skipped for fit, materials and compatibility questions.
• Barcodes as GTINs. The gtin in Product markup comes from the Shopify barcode field, and is left out for products with no real barcode. A SKU pasted into gtin is an invalid identifier, so the product cannot be matched to the same product listed elsewhere.
• Ratings only with reviews. aggregateRating only on products with published reviews. Review apps emit rating blocks for products with zero reviews, which makes the markup invalid and discounts the whole Product block.
• One brand name. The same brand string in the Organization markup, the Product markup, the footer and the contact page. A name written three ways gets merged with a similarly named business.

Shopify’s own list of the fields AI platforms consider adds type, vendor, collections, tags and variant option names (Optimizing your products for AI platforms), and Google asks that structured data match the visible text. Check again after launch: the checklist re-runs the structured data and JavaScript-off checks on day 30, because a new review or upsell app can add a second Product block or move the price into a script.

Store policies and the Knowledge Base app

Shopify asks for complete, current store policies, so that when a shopper asks about returns, “the AI agent can reference the correct information” (Optimizing your products for AI platforms). Its Knowledge Base app lets you review the automatically generated facts about your store that AI platforms use, see the questions shoppers commonly ask, and add answers of your own.

Two things you do not need to buy. Every Shopify store already serves /agents.md and /llms.txt with the store’s name, policies and sitemap, and Shopify says “you don’t need a third-party app” for them (Shopify Help). Google says you need no new “AI text files” for its AI features (Google Search Central). Our check shows whether a store has an llms.txt and never counts it against you: nice to have, not a ranking factor.

Step 3: answer shoppers’ questions on a page, not in a chat widget

Start from the questions your support team already answers before people buy. The answers have to sit where a crawler can read them. The checklist’s design and message pack asks for the ten questions support answers most often before purchase, each on a crawlable page as a question heading with the answer in plain text below it, “not inside a chat widget or a JavaScript accordion”. Answers that live only inside a chat script cannot be pulled out of the page by anything.

The Knowledge Base app’s list of common questions is a good place to find your ten. A price can be the answer too: on the site we built for Tipsy Pickle, the court and simulator rates are listed by time slot before anyone books.

Step 4: be mentioned by others

Assistants lean on other people’s sites. A 2025 preprint comparing AI search with Google found “a systematic and overwhelming bias towards Earned media”, meaning third-party sources, over brand-owned content (Chen et al. 2025).

So the slow work counts: reviews on sites shoppers trust, the stockists that list you, the press and roundups in your niche. Keep your name and facts the same everywhere. No markup replaces this step.

If Shopify Catalog leaves your products out

Shopify does not allow some categories in Catalog, among them “products that are age-restricted or regulated in many regions” (Requirements for being included in Shopify Catalog). The list includes alcohol; prescription and medical products; services and product add-ons, from professional services to warranties and subscription services; event tickets; and tobacco, CBD supplements, weapons, adult products and more. Shopify classifies each product from your product information, so the result “can differ from the product category that you set”, and the list “can change over time at Shopify’s discretion”. OpenAI’s own product rules apply on top: Shopify notes that products on OpenAI’s prohibited list do not show in ChatGPT’s product results (Selling on ChatGPT).

We have built for stores in and around these categories. Merican sells craft light beer, and every visitor passes a 21+ check before seeing a can; we did the full Shopify development (Merican). NEFT Vodka’s brand site, which we built in Framer, sells nothing itself: its Shop links send buyers to the brand’s Shopify storefront, which handles the alcohol checkout (NEFT Vodka). For ASPN Labs, a US concierge health service, some therapies could only be sold after an age and identity check or a prescription, and we built those checks into the path to checkout (ASPN Labs). At Tipsy Pickle, court and simulator bookings run on a custom booking platform we built beside the Shopify site (Tipsy Pickle).

For a store like these, look at Sales channels > Agentic before counting on the channel, and put the effort into Steps 1 to 4, which do not depend on Catalog. Say plainly on the page what the product is, what it costs and who may buy it. If the store opens on an age gate, do not guess what a crawler sees behind it: run the AI visibility check or the store checkup on the home page and a product page, and read what comes back.

How to tell whether it is working

Orders and referrals you can see

On Shopify, orders from AI channels show in the admin “with channel or referrer attribution” (Shopify Help). On the Agentic plan, you see ChatGPT orders only in your own store’s analytics (Shopify Help). In GA4, the default channel group has an AI Assistant channel for visits “from sources like ChatGPT, Gemini, Deepseek, Copilot, or Grok”, assigned when the referrer matches GA4’s list of assistants; it excludes Google’s AI Overviews and AI Mode (Default channel group).

Why ChatGPT visitors can show up as direct traffic

A customer says they found you through ChatGPT, and analytics says direct. GA4 files a visit as direct when it “doesn’t have a clear referral source”, and names the usual causes: links without UTM tags, redirects that strip them, an address typed or pasted into the browser, and ad blockers (GA4 Help). A shopper who reads about you in an answer and later types your address into the browser is, to GA4, a direct visit. Read the AI Assistant channel as a floor, not the total, and ask new customers where they heard of you.

Reports from Google and Bing, and a fixed prompt panel

Search Console’s Generative AI performance report shows how often links to your pages appeared in AI Overviews and AI Mode, by page, device and country; Google says it had reached all websites by August 31, 2026, and a site with too few impressions may not see it yet (Search Console Help). Bing Webmaster Tools’ AI Performance, introduced as a public preview in February 2026, shows how often your pages are cited in Microsoft Copilot, AI summaries in Bing and some partner integrations (Bing Webmaster Blog).

To see what the assistants say, keep a fixed prompt panel: a handful of questions a buyer would ask (one by category, one by problem, one with your brand name), asked the same way every month, each in a new chat, with web search on and memory off, a few runs each. Save the answers word for word and note whether you are named and whether the facts are right. The same study found that answers shift with phrasing, and by different amounts in each AI search service, so one screenshot proves nothing either way.

What nobody can promise

Each platform says it in its own words. Shopify: eligibility “doesn’t guarantee that a product displays in a specific AI answer”, and each AI channel decides (Shopify Help). OpenAI: “Placement is not guaranteed” (Searching the web with ChatGPT). Google: indexing and serving “isn’t guaranteed” (Google Search Central). Our own check says it too: “Being let in comes first, and it does not mean you will be named.” Be wary of anyone who sells a guaranteed place in ChatGPT answers.

What we do covers Steps 1 to 3, as SEO and AI visibility, from $400 a month: technical fixes and product data that search engines and AI assistants can read, content your buyers actually search for, and a monthly report on rankings and AI answers. It runs three months to start, then month to month with 30 days’ notice, and if it is not paying off, we will tell you to stop. To start smaller, run the free AI visibility check on your home page, or book a store review: a free 15-minute call, then written notes.