How to get your product recommended in ChatGPT Shopping

The common mistake is treating ChatGPT Shopping like a normal search result. You polish a product page and wait for it to appear, while the product data stays buried in scripts, feeds or a checkout flow. ChatGPT can name a competitor's version instead, or quote a price your feed retired three months ago.

Shopping answers may draw on product pages, structured data and merchant information. OpenAI's commerce documentation now describes a product feed: merchants share a regularly refreshed CSV or JSON file so ChatGPT can surface their products in search and shopping. OpenAI does document that OAI-SearchBot is used to surface websites in ChatGPT search features, and that sites opted out of it will not appear in search answers. In our readiness audit, Product or SoftwareApplication schema failed on 59 of 121 sites.

Product data gives answers substance

A shopping answer needs usable product facts. It may need the product name, what it does, who it suits, its price and how it differs from alternatives.

Treat your page, schema and merchant feed as connected sources. Explain the product in ordinary language on the page, then keep the structured records consistent with that explanation.

The readiness checks show the gap. Product or SoftwareApplication schema passed on 60 of 121 sites. A plain statement of what the product is, who it is for, its price and what it is against passed on 20 sites, failed on 43 and warned on 56.

The checks do not always total 121 because some sites returned no result for a particular check. Read each result as a measure of that check, not as a complete classification of every site.

Give the assistant clean product facts, then explain those facts in ordinary language on the page.

Make the product page answerable

Start with the page a buyer would want to cite. Its first visible section should identify the product and its use, rather than making the reader decode a slogan.

A useful page can answer these questions without a second visit:

1. Name the product in one plain sentence.
2. State who it is for or which use case it serves.
3. Show the price or link directly to a pricing page.
4. Explain the main difference between this product and a relevant alternative.
5. Put important product facts in the rendered page, rather than revealing them only after a script runs.
6. Make the Product or SoftwareApplication markup match the visible product name, description and offer details.
7. Answer common buyer questions in language a person would use.

This is not a request to turn every product page into a long article. Remove the small acts of interpretation that make extraction unreliable.

Our scan found that a price was visible or a pricing page was linked on 90 of 121 sites. One clear H1 passed on 101, and content was present without JavaScript on 92. Basic checks still matter.

What OpenAI documents about product feeds

OpenAI's commerce documentation says merchants share structured product data with OpenAI so ChatGPT can accurately surface their products in search and shopping experiences. The feed is a secure, regularly refreshed CSV or JSON file with identifiers, descriptions, pricing, inventory, media and fulfillment options.

The same page separates two kinds of field. Required fields make sure price and availability display correctly. Recommended attributes, such as rich media, reviews and performance signals, are the ones OpenAI says improve ranking, relevance and user trust. Integration starts with an initial sample feed for validation, followed by daily snapshots.

The documentation describes the feed. It does not promise that a product in a feed will be recommended, and the page still has to explain why the product fits the buyer's question.

Keep feed and page facts aligned

A merchant feed should not tell one story while the product page tells another. Keep the product name, description, availability, price and destination consistent wherever those fields exist. The assistant has no reason to prefer the version you meant.

Use the product page as the explanation and your merchant feed as a clean record of product facts. That keeps the page readable without forcing it to carry every structured field in visible prose.

A practical handoff has two parts. The page says who the product suits, what it replaces or complements and why its key features matter. The schema identifies the product and its offer.

The feed carries the merchant data required by your shopping setup. Your internal process checks that changes reach the page, schema and feed, so an old price does not survive in one place.

This is where the free AI readiness check helps. It cannot certify that ChatGPT will recommend a product, but it can show obvious access, markup and page clarity problems before you spend time on promotion.

Earn the pages around your product

Your own product page is one possible source. In our scan, listicles and roundups supplied 67.5% of recorded citations, or 30,585 citations. Competitor product pages supplied another 24.4%, or 11,052 citations. Together, those page types made up 91.9% of citations. The figures come from CiteGraph's research scan.

That changes the job. You need a sound product page, but you also need to understand the pages that help assistants compare products.

Find the buying questions where your product belongs, then make the evidence easy to verify. A comparison page, independent roundup or useful category guide may be more likely to enter an answer than your homepage.

The data argues against spending the whole budget on your own domain. The product's own site accounted for 1.5% of recorded citations, while community threads accounted for 3.4%. This is a sample, not a universal law, so treat it as a signal of what is common rather than a guarantee.

Use the weekly research on who owns AI answers to watch page types and source patterns in the sample. Then use the free AI visibility checker to test whether your product appears for the buyer prompts that matter to you.

Access still comes before selection

Structured product data cannot help if the relevant pages cannot be fetched. OpenAI says that OAI-SearchBot is used to surface websites in ChatGPT search features, and that sites opted out of OAI-SearchBot will not be shown in ChatGPT search answers, although they can still appear as navigational links. The guidance is in OpenAI's bot documentation.

Check robots.txt before diagnosing a content problem. In the readiness scan, AI crawlers were allowed in robots.txt on 116 of 121 sites, so five sites failed that check. OpenAI says it can take about 24 hours from a robots.txt update for its systems to adjust.

A sitemap helps discovery, although it is not a recommendation signal by itself. A sitemap was published on 100 of 121 sites. Canonical and Open Graph checks passed on 92, with warnings on 27.

Here is the argument against this approach: perhaps this is ordinary ecommerce SEO with a new label. A strong product page, a crawlable site and accurate feeds are sensible work anyway, so a separate ChatGPT Shopping process could add bureaucracy without adding visibility.

That objection is partly right. You should not create a second website for an assistant. The useful difference is where you put the emphasis: check product schema and merchant data alongside page clarity, rather than assuming content work covers them. The schema result above shows why that check is worth making.

Measure recommendation paths, not hope

A ranking report is too narrow for ChatGPT Shopping. Check whether the product is named, whether it appears in a shopping answer, which page supports the answer and whether the cited facts match the current offer.

Test prompts that reflect real buying decisions. Include category searches, use case searches, comparisons, alternatives and price-constrained requests. Save the exact prompt, the products named, the cited URLs and the facts used, then repeat the same set after meaningful changes.

Before and after, one product page can become clearer without becoming longer. Before: a page for a project management app opens with the slogan “Work better together”, hides the monthly price below a form, lists task boards and time tracking without naming small agencies as the buyer, and has incomplete Product schema. After: the H1 says “Project management software for small agencies”, the opening sentence names that buyer, the price is visible, a short comparison explains when a spreadsheet is the weaker alternative, and the structured product data matches the page.

None of that guarantees a recommendation.

The useful next question is: which product facts can ChatGPT find right now, and where does the answer live? The answer lives in the readiness checks, the free AI visibility checker and the source pages returned for your own shopping prompts.