Skip to content

AI Search Optimization (AEO) · focused work

ChatGPT shopping visibility

When a shopper asks ChatGPT what to buy, it answers with a shortlist: product cards with prices, a summary of reviews and a link to a store. The stores that appear are the ones whose product data the assistant can read, trust and match to the question. That is a different problem from ranking in Google, with its own access rules, data routes and measurement.

This is the right work if

  • The store blocks every AI crawler and has not decided whether search assistants should be allowed
  • Analytics shows referrals from chatgpt.com landing on dead pages or the homepage
  • Product pages present specifications in images, PDFs or tabs that load after the page renders
  • The store sells a category where shoppers ask an assistant for a shortlist before they visit any site
  • Nobody has typed the store's own buying questions into ChatGPT to see who appears

What it is

The work behind ChatGPT shopping visibility

ChatGPT shopping visibility is the work of making a store's products eligible to appear when ChatGPT answers a buying question, and of measuring what those appearances send. ChatGPT builds its product answers from web pages its search crawler has read, from product data supplied by merchants and data partners, and from review content it can attribute. The crawler route needs robots.txt to allow OAI-SearchBot, pages that render product facts in HTML and stable URLs. The feed route needs a product feed in the format OpenAI accepts, kept accurate on price and availability.

The content route is closer to conventional SEO: product pages that state what the item is, who it suits and its specifications in plain text, plus guide pages that explain how to choose within the category, because the assistant reads those to decide which products fit a stated need. Around all of it sits measurement: referral traffic from chatgpt.com in analytics, landing pages and revenue from that source, and a periodic check of the store's own prompts. The aim is presence in the shortlist, never a promised placement.

How it is done

The work, in order

What changes

  • The store's products can appear in ChatGPT's shopping answers because its data reaches the assistant by several routes
  • Referral traffic from ChatGPT is measured with landing pages and revenue instead of being lumped into direct
  • Product pages state facts an assistant can quote, which also helps a human shopper decide
  • The store knows, for its own buying questions, which products and stores appear and how that changes month to month
  1. Set crawler access deliberately

    Separate the crawlers by purpose: OAI-SearchBot fetches pages for search answers, ChatGPT-User fetches when a person asks, and GPTBot collects training data. Decide each one in robots.txt, and check that firewalls, bot managers and CDN rules match the file. A store that wants to appear must allow the search crawler.

  2. Test the store's own questions

    Write the twenty or so questions a shopper in the store's category would ask an assistant, and record what ChatGPT returns for each: which products, which stores and which sources it cites. This is the baseline everything else is measured against.

  3. Make product pages readable and specific

    Put name, brand, price, availability, key attributes, materials, sizing and shipping scope in the HTML as text, with Product markup that matches. Add a short statement of who the product suits and what it is not for; an assistant matching a stated need reads those lines first.

  4. Supply product data where OpenAI accepts it

    Where the store's platform and region are eligible, prepare a product feed to OpenAI's published specification with the same discipline as a Merchant Center feed: accurate prices, availability, identifiers and images. Where checkout inside ChatGPT is offered to merchants on the platform, weigh eligibility and margin before opting in.

  5. Write the choosing content

    Buying guides and comparisons within the store's categories, written from product knowledge, that explain which type suits which need. Assistants use pages like these to decide what to recommend, and they cite them. Each guide links to the products it describes so a reader who arrives can buy.

  6. Track and repeat

    Build a referral segment for chatgpt.com in analytics with landing pages, sessions and revenue, and repeat the question baseline monthly. Movement in the shortlist is the leading indicator; revenue from the referral segment decides how much more to invest.

Platform notes

Shopify

Shopify has built commerce integrations with OpenAI so eligible merchants' catalogs can surface, and in some regions transact, inside ChatGPT; check the store's eligibility and settings in the Shopify admin rather than assuming it is on.

WooCommerce

There is no built-in route, so product data reaches ChatGPT mainly through crawling; server-rendered product pages with complete Product markup carry the load, and any feed is prepared with a feed plugin or a custom export.

Questions

ChatGPT shopping visibility questions

By CartKernel · Last reviewed

Does allowing OAI-SearchBot mean OpenAI trains on the store's content?

No. OpenAI documents GPTBot as the training crawler and OAI-SearchBot as the crawler for search features, and they are controlled separately in robots.txt. A store can allow search fetching and disallow training, or the reverse.

What makes a product eligible for a ChatGPT shopping answer?

The assistant needs to have read the product, either from a crawlable page or from supplied data, and needs enough stated attributes to match it to the question. Products with vague names, missing prices or specifications locked in images are hard to match.

How long does it take for changes to show up in ChatGPT answers?

Crawled pages are refreshed on the crawler's own schedule, and supplied feeds update on theirs, so expect weeks rather than days between a page change and a changed answer. Judge trends over a quarter, not single prompts.

Should the store optimize for Perplexity and Gemini at the same time?

Yes, and most of the work carries across. Each assistant has its own crawler to allow and its own referral domain to track, but all of them read the same product pages, so a page written clearly for one is readable by the others.

Related work and answers

AI Search Optimization (AEO)AI Overviews for ecommerceAI Overviews for ecommerce: finding the buying queries that trigger them, making guide and product pages citable, and keeping feed data accurate.OpenAI Search Optimization (AEO)Product schema for AI searchProduct schema for AI search: one accurate JSON-LD block per page, complete merchant listing fields, honest variant modelling and validation that holds.OpenProduct Feed OptimizationProduct feed title optimizationProduct feed title optimization: a pattern per category built from real queries, generated from product attributes, kept in policy and measured properly.OpenProgrammatic SEOProgrammatic buying guidesProgrammatic buying guides: comparison tables built from catalog data, an editorial judgement in every guide, a publication gate and a refresh cycle.OpenAnswerHow does ChatGPT decide which products to recommend?ChatGPT recommends products from what it retrieves at answer time: crawled pages, merchant feeds and third-party reviews, filtered by the shopper's constraints.OpenAnswerShould an online store block AI crawlers?Most stores should allow the AI crawlers that answer shoppers and decide separately about training-only bots. Which user agents do what, and a robots policy.OpenAnswerHow do you track traffic from AI search?Assistant referrals show up in analytics by hostname, while Google's AI surfaces do not separate out. How to build the channel and read the indirect signals.OpenAnswerWhat do shoppers ask AI assistants before buying?Shoppers ask assistants long, conditional questions rather than keywords. The five question shapes that recur and what they require from your product data.Open

Find the leak.

A free Growth Analysis ranks what your store should fix first, by revenue at stake.