How ChatGPT Decides Which Products to Show
ChatGPT, Google AI Mode and Perplexity pull product answers from merchant feeds first and the open web second. Here is what each one reads, which crawlers to allow, and what changed in March 2026.
- By
- VendaHawk
- Published
- Updated
- Reading time
- 6 min read
ChatGPT shows products from two sources: a structured product feed supplied by the merchant (or by Shopify on the merchant's behalf) and pages its search crawler, OAI-SearchBot, has indexed. Google AI Mode and Perplexity work the same way, with Merchant Center and Shopify Catalog as their feeds. Sellers control the feed fields, the product page text, and the robots.txt rules that let the crawlers in.
Feeds come first, markup second
Every major AI shopping surface ingests a structured feed as its primary product source. OpenAI's commerce guide says merchants provide a "secure, regularly refreshed feed (CSV or JSON)" carrying "identifiers, descriptions, pricing, inventory, media, and fulfillment options", and that recommended attributes such as rich media and reviews "improve ranking, relevance, and user trust" (OpenAI key concepts). Google is even more direct about on-page markup: "There's also no special schema.org structured data that you need to add" to appear in AI Overviews or AI Mode (Google AI features).
That does not make JSON-LD useless. It is a consistency signal that lets engines reconcile the page with the feed. But it is not the entry ticket. The feed is.
| Surface | Primary product source | Search crawler | Where the buyer checks out (Oct 2026) |
|---|---|---|---|
| ChatGPT search and shopping | OpenAI product feed (approved partners) or Shopify Catalog; open web | OAI-SearchBot | Merchant's own checkout; Instant Checkout moved to Apps |
| Google AI Mode and AI Overviews | Merchant Center feed into the Shopping Graph; indexed pages | Googlebot (Storebot-Google for Shopping) | Merchant site; UCP "Buy" button for select merchants |
| Perplexity | Shopify Catalog; Merchant Program; PerplexityBot index | PerplexityBot | Instant Buy with PayPal for US users; merchant remains the seller |
| Microsoft Copilot | Microsoft Merchant Center feed plus web; Shopify auto-enrolled | bingbot | Copilot Checkout (US) |
What ChatGPT reads from a feed
OpenAI's feed spec requires item_id, title (up to 150 characters), description (up to 5,000 characters of plain text), url, brand, seller_name, image_url, availability and price. Optional fields that carry weight include gtin, mpn, color, size, material, additional_image_urls, shipping, accepts_returns, review_count and star_rating (OpenAI product feed spec). The enable_search flag defaults to true; enable_checkout defaults to false.
OpenAI's own best practices are short: "use concise, factual copy", keep a stable parent id with unique variant ids, give each variant its own title, URL, media, availability and price, and tag feed URLs with utm_medium=feed so you can see the traffic (OpenAI best practices). OpenAI recommends one full file upload a day plus API upserts during the day.
Feed onboarding is "currently available to approved partners" through chatgpt.com/merchants (OpenAI get started). Most small and mid-size sellers reach ChatGPT through Shopify instead.
The Shopify path and the March 2026 checkout change
On 6 March 2026 OpenAI said "Instant Checkout is moving to Apps, where purchases can happen more seamlessly", and that ChatGPT would prioritise search and product discovery with the Agentic Commerce Protocol as the connective layer (Digital Commerce 360). In practice, ChatGPT now finds the product and sends the buyer to the merchant's own checkout.
Shopify's Agentic Storefronts channel is "active by default for eligible stores". ChatGPT buyers "buy on your online store checkout, in a ChatGPT in-app browser or in a new tab". Shopify Catalog sends title, description, options, images, price and availability, "updated continuously", and Catalog Mapping can read metafields, metaobjects and tag prefixes. You can opt out per channel under Sales channels > Agentic, and setting a product to Unlisted hides it from every AI channel and from search (Shopify Agentic Storefronts).
Google AI Mode: be indexed, be snippet-eligible, keep the feed current
Google's rule for AI Overviews and AI Mode is that pages "must be indexed and eligible to be shown in Google Search with a snippet", with "no additional requirements". Googlebot's robots.txt rules are the control; nosnippet, data-nosnippet, max-snippet and noindex limit what AI features can show (Google AI features).
Two tokens confuse people. Google-Extended is a standalone product token that governs training of future Gemini models and grounding in Gemini apps; it "does not impact a site's inclusion in Google Search nor is it used as a ranking signal" (Google common crawlers). Blocking it does nothing to AI Mode.
For product data, Merchant Center free listings need id, title, link, image_link, price, description and availability (Merchant Center requirements). Since May 2026, Merchant Center's AI performance insights report your share of voice in AI Mode and AI Overviews, the competitor average, top terms and the popular attributes "missing from your product data" for English queries in the US, CA, AU, IN and NZ (AI performance insights).
Perplexity: Shopify data, a merchant programme, and two bots
Perplexity's shopping answers draw on Shopify-powered stores that sell and ship to the US and on its free Merchant Program, which was built for retailers "to share their product specs" (Perplexity blog). Instant Buy is now "available to all users in the United States" with PayPal checkout, and purchases "are from the Merchant directly".
PerplexityBot indexes pages for search results and "is not used to crawl content for AI foundation models". Perplexity-User fetches a page when a user asks and "generally ignores robots.txt rules" (Perplexity bots).
The crawler tokens that matter
Blocking the wrong bot is the cheapest way to disappear. "Sites that are opted out of OAI-SearchBot will not be shown in ChatGPT search answers", while GPTBot is training-only and disallowing it does not affect search (OpenAI bots). Robots changes take about 24 hours to propagate.
Allow these search and user-fetch tokens: OAI-SearchBot, ChatGPT-User, Googlebot, Storebot-Google, PerplexityBot, Perplexity-User, Claude-SearchBot, Claude-User, Amzn-SearchBot, Amzn-User, Applebot, bingbot. Decide separately, as a brand policy, on training tokens: GPTBot, Google-Extended, ClaudeBot, Amazonbot, Applebot-Extended, meta-externalagent, CCBot. Check your CDN too: some "block AI bots" toggles block the search bots as well.
What actually gets a page cited
Ahrefs' study of 1.4 million prompts found that cited URLs had a title-to-prompt similarity of 0.602 against 0.484 for non-cited pages, and that natural-language slugs were cited 89.8% of the time against 81.1% (Ahrefs). AI-cited pages are also fresher, averaging 1,064 days old against 1,432 for organic results (Ahrefs freshness). Ranking well in Google is not the same thing: only about 12% of ChatGPT, Gemini and Copilot citations rank in Google's top ten (Ahrefs overlap).
For a product page, that means a title that matches how shoppers phrase the question, a 40 to 60 word answer-first paragraph, a spec table whose numbers match the feed, and a refresh at least quarterly.
Frequently asked questions
Does ChatGPT read my Product schema markup?
It can crawl it with OAI-SearchBot, but the feed is the primary product source. Markup helps engines confirm that the page, the feed and the visible text agree. Google states outright that no special structured data is required for AI features. Treat JSON-LD as a consistency check, not the way in.
Can I still sell inside ChatGPT after the March 2026 change?
Buying now happens on your own checkout. OpenAI moved Instant Checkout into ChatGPT Apps on 6 March 2026 and refocused ChatGPT on discovery. Shopify stores open the merchant's checkout in a ChatGPT in-app browser or a new tab, with the merchant as seller of record.
Should I block GPTBot?
That is a training decision, not a visibility decision. OpenAI says disallowing GPTBot does not affect ChatGPT search. The token you must not block is OAI-SearchBot. Set your training policy once, document it, and keep the search and user-fetch bots allowed.
Do I need an llms.txt file?
It takes minutes and Shopify emits one automatically, but no engine officially consumes it for ranking or discovery, and Google says you do not need new machine-readable files to appear in AI features. Ship it, and do not expect measurable impact from it.
How does Perplexity get product data from a non-Shopify store?
Through its PerplexityBot index of your product pages and through the Merchant Program signup. Keep product pages crawlable, keep prices and availability current in the visible text, and give each product one canonical, snippet-eligible page.
Sources
- https://developers.openai.com/commerce/guides/key-concepts.md
- https://developers.openai.com/commerce/specs/file-upload/products.md
- https://developers.openai.com/commerce/guides/best-practices.md
- https://developers.openai.com/commerce/guides/get-started.md
- https://developers.openai.com/api/docs/bots
- https://www.digitalcommerce360.com/2026/03/06/openai-shifts-checkout-plans-agentic-commerce-strategy/
- https://help.shopify.com/en/manual/online-sales-channels/agentic-storefronts
- https://developers.google.com/search/docs/appearance/ai-features
- https://developers.google.com/search/docs/crawling-indexing/google-common-crawlers
- https://support.google.com/merchants/answer/13889434
- https://support.google.com/merchants/answer/17200695
- https://www.perplexity.ai/hub/blog/shop-like-a-pro
- https://docs.perplexity.ai/guides/bots
- https://ahrefs.com/blog/why-chatgpt-cites-pages/
- https://ahrefs.com/blog/do-ai-assistants-prefer-to-cite-fresh-content/
- https://ahrefs.com/blog/ai-search-overlap/
Questions this article answers
Does ChatGPT read my Product schema markup?
It can crawl it with OAI-SearchBot, but the feed is the primary product source. Markup helps engines confirm that the page, the feed and the visible text agree. Google states outright that no special structured data is required for AI features. Treat JSON-LD as a consistency check, not the way in.
Can I still sell inside ChatGPT after the March 2026 change?
Buying now happens on your own checkout. OpenAI moved Instant Checkout into ChatGPT Apps on 6 March 2026 and refocused ChatGPT on discovery. Shopify stores open the merchant's checkout in a ChatGPT in-app browser or a new tab, with the merchant as seller of record.
Should I block GPTBot?
That is a training decision, not a visibility decision. OpenAI says disallowing GPTBot does not affect ChatGPT search. The token you must not block is OAI-SearchBot. Set your training policy once, document it, and keep the search and user-fetch bots allowed.
Do I need an llms.txt file?
It takes minutes and Shopify emits one automatically, but no engine officially consumes it for ranking or discovery, and Google says you do not need new machine-readable files to appear in AI features. Ship it, and do not expect measurable impact from it.
How does Perplexity get product data from a non-Shopify store?
Through its PerplexityBot index of your product pages and through the Merchant Program signup. Keep product pages crawlable, keep prices and availability current in the visible text, and give each product one canonical, snippet-eligible page.