ChatGPT shopping optimization: A checklist for store owners

The checklist

1. Let OpenAI's search crawler in

OpenAI documents OAI-SearchBot as the crawler "used to surface websites in search results in ChatGPT's search features". If it cannot fetch your product pages, ChatGPT cannot read them directly.

• Check robots.txt for a rule that blocks OAI-SearchBot, or a broad rule under User-agent: * that blocks product paths.
• Check your CDN or host's bot protection. Store platforms and security services sometimes challenge unknown bots, which silently drops crawler requests.
• Remember GPTBot is a separate crawler for model training. Blocking it does not block search, and allowing it does not help search.

OpenAI says a robots.txt change can take about 24 hours to take effect.

The frame at the top is that check on our own site, which sells software, not products: in the 30 days to 2026-10-04, OAI-SearchBot was seen on all 30 of our most-fetched pages, and /pricing was fetched by all three OpenAI agents. Run the same check on your product and category pages; the crawler log shows it, robots.txt alone does not.

2. Make product pages readable without JavaScript

Crawlers fetch HTML. If your product title, price, description and availability only appear after scripts run, a crawler may see an empty shell. Most hosted store platforms render product pages on the server, but themes and apps can move content into scripts. Fetch a product page with a plain HTTP client and check the important text is there.

3. Keep product data consistent everywhere

ChatGPT uses structured data from third-party providers. That means your price, availability and description live in several places you do not fully control: marketplaces, comparison and price sites, feeds you syndicate. When they disagree, the engine has to pick one.

• Add Product structured data to your product pages with name, price, currency, availability, brand and an identifier such as GTIN where you have one.
• Make the visible price and the structured price identical.
• Audit the listings you syndicate to, and fix stale ones first for your best sellers.

4. Write product copy that matches how shoppers ask

Shoppers describe needs, not product names. "For flat feet", "for a small kitchen", "for beginners", "quiet". If your description lists specifications but never says who the product is for, there is nothing for a needs-based question to match.

For each important product, the description should state in plain sentences: who it is for, the problem it solves, what makes it different from the obvious alternative, and the facts a shopper would check (size, material, compatibility, care). One clear paragraph beats a wall of bullet points.

5. Earn reviews and mentions where ChatGPT reads

OpenAI names reviews and third-party content as inputs. For product questions, the third-party pages that decide answers are usually review sites, buying guides, roundups, forums and videos. The work here is the same as classic PR, aimed at the pages that already appear in answers:

• Ask customers for reviews on the platforms your category uses.
• Send products to the editors of buying guides that ChatGPT cites for your category.
• Answer real questions in communities where your products are discussed, openly as the brand.

6. Consider OpenAI's merchant programme

OpenAI runs a merchant programme through which approved merchants can share a product feed directly. If you sell at scale, apply. It does not replace the steps above, because product selection still depends on relevance and third-party content.

7. Keep collection and category pages useful

Collections such as "running shoes for flat feet" are pages that match shopper questions directly. A collection with an introduction that explains who it is for, and how to choose within it, is a strong candidate for both Google and ChatGPT. A collection that is only a grid of thumbnails gives an engine nothing to quote.

Which product pages should you fix first?

A store with hundreds of products cannot rewrite them all at once, and does not need to. Order the work by where an improvement can change an answer:

1. Products that already appear in some answers. They are in the candidate set; a clearer description and consistent data can move them from occasional to regular.
2. Best sellers in categories where competitors are named and you are not. These have the revenue at stake and a known gap.
3. Collections that match common shopper needs. One good collection page can serve dozens of questions.
4. Products with wrong facts in answers. Fix the source of any wrong price or specification immediately, whatever the product's rank.

Everything else can follow a steady rhythm, a handful of products a week, measured as you go. Rewriting a whole catalogue in one pass makes it impossible to tell which changes helped.

What should you measure?

Shopping visibility is measured the same way as any AI visibility: read the answers.

1. Write 20 to 50 shopping questions in shoppers' words, covering your main categories and needs.
2. Ask ChatGPT each one several times, over several days. Product results vary between samples.
3. Record for each answer: whether your product appeared, which competitors appeared, which merchants were shown and which sources were cited.
4. Count, with sample sizes. "Our product appeared in 6 of 27 answers" is a reading.
5. Re-read after each change and compare against the weeks before it.

Also check what ChatGPT says about your products when it does show them. A wrong price or a wrong compatibility claim costs sales. See fixing what AI gets wrong about your brand.

What should you not do?

• Do not stuff product titles with need phrases. "Best quiet coffee grinder for small kitchens 2026" as a product title reads as spam to shoppers and to engines. Put the needs in the description.
• Do not hide prices or availability. A product without a clear price is hard to recommend for a budget question.
• Do not pay for reviews. Beyond the platform rules it breaks, it puts your brand's name next to the word "fake" in the sources engines read.
• Do not treat ChatGPT ads as a shortcut. Ads are a separate, labelled placement. They do not change which products the answer itself recommends. See how to prepare for ChatGPT ads.

How long does it take?

OpenAI does not publish how often it re-reads product data, so no one can promise a date. Some steps act quickly: a crawler block removed takes effect within about a day of OpenAI reading your robots.txt. Others are slow by nature: reviews and buying-guide mentions accumulate over months. Measure weekly, judge monthly.

Is this different from Google shopping?

The inputs overlap: product data, reviews, accurate pages. The difference is in how the shopper asks. Google Shopping matches a product query to listings. ChatGPT reads a described need, often a long one with constraints, and recommends a few products with reasons. That rewards product copy that explains fit, and third-party content that discusses products in context. A store that does both is ready for either.

How does Attensira help with ChatGPT shopping?

For a store, Attensira tracks the shopping questions your customers ask, on ChatGPT and the other engines your plan reads, and shows which products and merchants each answer named and which pages it relied on. With the Shopify connector, the crew can rewrite the copy on your products and collections and add new blog articles. It changes three fields only, the description, the SEO title and the SEO description, and never touches prices, variants, images, orders or customers. Every change waits in your inbox for approval; when Shopify has no draft version of a field, the proposal is kept beside the product as an attensira metafield you can see in your own admin. New articles are always added unpublished, so publishing stays yours. After a change ships, the same questions are re-read and reported with their sample size.

That is the service of an SEO agency at the price of a tool, for a store. See Attensira for ecommerce, or run the free audit on your store's domain.