Get your products recommended by ChatGPT

How ChatGPT decides which products to recommend

ChatGPT doesn't keyword-match the way old search did. When someone asks “what's a good standing desk for a small apartment under $300?”, it weighs relevance to that specific intent, the structured facts on your product pages (price, availability, specs), review signals and authority, and how well your copy matches the shopper's real constraints. Buyers ask AI long, situational questions — far more detail than a Google search — so the products that win are the ones whose pages speak in those same situational terms.

Step 1: Become eligible (crawler + feed)

Let the right crawlers in

ChatGPT uses OAI-SearchBot for its search and shopping surfaces and GPTBot for broader retrieval. If either is disallowed in /robots.txt— or your CDN quietly blocks them — your catalog can't appear, full stop. This is the single most common reason an otherwise great store is invisible. See the GPTBot referencefor the exact directives, and double-check that bot-management rules aren't catching these user-agents.

Get your catalog into the shopping surface

Product carousels in ChatGPT are fed by structured product data, not just the open web crawl. If you sell on Shopify or Etsy, your catalog is largely integrated already — little or no extra setup. Otherwise you submit a product feed (titles, descriptions, prices, availability, identifiers) through OpenAI's merchant application so products are eligible for discovery. Keep that feed accurate and current; stale price or stock data is worse than none, because it gets you shown and then filtered out.

Step 2: Get chosen (product-page copy)

Eligibility gets you into the pool. Your product page decides whether you're recommended. Four moves do most of the work:

1. Lead answer-first.Open with what the product is, who it's for, and the one-line reason to buy — before the brand story. A model may stop reading before it reaches a buried spec.
2. Write in use cases, not just specs.Swap “10,000mAh battery” for “about 48 hours of power for off-grid camping.” Buyers ask in scenarios; match the scenario.
3. Make facts extractable. Put price, sizes, materials, compatibility and shipping in plain text and a comparison-friendly format, not locked in an image or a script-rendered widget.
4. Show reviews as text.Genuine, crawlable reviews give the model the language buyers use and the problems your product solves — strong context for a recommendation.

Step 3: Add Product structured data

Mark up each page with schema.org Product (name, brand, description, Offer with price and availability) and AggregateRatingwhen you have real reviews. Structured data turns prose into facts a model can attribute with confidence, and it is what shopping surfaces and autonomous buying agents read first. The same principle drives the rest of AEO — see schema/JSON-LD for AI assistants.

Step 4: Don't hide the page behind JavaScript

Most AI crawlers fetch raw HTML and don't wait for client-side rendering. If your price, variants or description only appear after a framework hydrates, the bot sees an empty shell. Quick test: view-source the page. If the buying facts a human reads aren't in that raw HTML, neither ChatGPT nor an agent shopping on a buyer's behalf will use them.

How long it takes

Feed and product-data improvements tend to surface in AI shopping results within a few weeks as the index refreshes. Building external authority — reviews, mentions, being listed in roundups — is the slower 3–6 month layer. Fix access and product data first; it's the fastest, highest-leverage work.

FAQ

Check whether AI can read your product pages per agent: run a free audit, or see how the AI Shopping add-on surfaces your catalog in ChatGPT, Gemini and Perplexity.