A retail or DTC brand gets named in an AI shopping answer when the assistant finds product information that matches what the shopper asked for and decides to use it. That decision belongs to the AI system, and no brand controls it. What a brand does control is whether its product pages, product data, and policy pages state fit, materials, ingredients, price, delivery, and returns in plain text, the same way everywhere, so any system that reads them has accurate facts to pass along.
Shopping questions arrive loaded with conditions. A shopper might ask for a linen shirt that does not need ironing, costs under a set amount, ships to Canada, and can be sent back by mail. Every one of those conditions gives an answer a reason to include one product and leave out another, and a product page that answers none of them leaves the gaps to reviews, marketplaces, and resellers.
This page is content planning for brand owners and ecommerce leads, not legal advice. The federal rules that shape product claims, reviews, and origin statements are summarized in our guide to content for retail, DTC and beauty brands; take the specifics to your counsel.
What OpenAI and Google say about shopping answers
OpenAI's help article on shopping in ChatGPT search says a product appears in its carousel "when ChatGPT perceives it to be relevant to your intent," and that product results are selected independently and are not ads. Among the things it says ChatGPT considers is structured metadata from first-party and third-party providers, such as price and product description, and it says ChatGPT may also show model-generated review summaries based on reviews from public websites. The same article says reviews and ratings are not verified by OpenAI, and that when a shopper opens the list of merchants for a product, merchants are ranked on factors like availability, price, and quality. It also points merchants to OpenAI's developer documentation to apply for direct product feed access.
OpenAI's publisher FAQ adds the access condition: public sites are eligible for ChatGPT search, and a site's content shows up in summaries and snippets only if it is not blocking OAI-SearchBot. Google's AI Overviews and AI Mode documentation says there are no additional requirements or special optimizations for those features, that a page used as a supporting link has to be in the index and able to appear with a snippet, and it lists keeping Merchant Center and Business Profile information up to date among its suggestions. Neither company publishes how one product earns a place over another in a given answer.
Your product facts live in more than one place
For most brands, a product is described in at least three places: the product page, the product data sent to feeds and marketplaces, and the structured data in the page code. Google's merchant listing documentation says Product markup can make a page eligible for merchant listing experiences, including the shopping knowledge panel, Google Images, popular product results, and product snippets. Our working hypothesis is that when those sources disagree, a summary built from them may repeat whichever version it happens to read, so the planning goal is a single version of each fact.
| Fact | On the product page | In the feed or markup | Who confirms it |
|---|---|---|---|
| Price and sale price | The visible price, with sale end dates | Price in the feed and in the Offer markup | Ecommerce manager |
| Size range and fit | A fit note for each style | Size values and variants | Merchandising lead |
| Materials or ingredients | Plain text, not only inside an image | Material field or description | Product development or regulatory reviewer |
| Shipping and returns | A short policy summary near the price | Shipping and return settings | Operations and customer service |
| Availability | Stock status and restock timing | Availability value in the feed | Inventory planner |
A useful working session puts the ecommerce manager and the customer service lead in front of one screen with the product page, the feed export, and a marketplace listing open side by side. Ask them, "Where do these three disagree about price, fabric, size range, or the return window?" Each disagreement they find is a fact the brand is currently telling the world two ways.
Review summaries are written by someone else
Because OpenAI describes its review summaries as generated from public websites, the reviews on your own site and on retailer and marketplace pages can shape how a product gets described. Shoppers answer each other's questions there: "runs small," "darker in person," "the lid is not oven safe." A review summary is not a product answer. When the same question keeps surfacing in reviews, the stronger move is to answer it on the product page in the brand's own words, with the facts behind it, so reviews are not the only written source on the point.
That does not mean steering or filtering reviews. FTC rules govern how reviews are collected and shown, which we cover in how reviews and user photos should be used under FTC rules. The point here is narrower: read your reviews as a list of questions your pages have not yet answered.
A fictional sleepwear brand runs its own searches
Take a fictional DTC sleepwear brand in Oregon. Its founder asks an assistant for breathable pajamas for hot sleepers that survive the dryer and come with free returns. The response names two competitors and links pages that state fabric weight, fiber content, dryer guidance, and the return window, while the fictional brand's own page says "cloud-soft comfort for your best night." Rewording the prompt is not the fix. The people drafting those pages should never have to invent the fabric weight, the dryer guidance, or the return terms, because the tech pack, the wash testing notes, and the returns policy already hold them; the writer's job is to get those facts from the people who own them and set them down in plain text.
Keep a record when you run searches like this: the date, the exact wording, any location setting, the full response, and every source it linked. A one-off answer is an anecdote, and repeating the same questions on a schedule is what turns anecdotes into a pattern you can act on. Separately, our free AI visibility check is an eight-question self-check about your own website that takes about two minutes. It does not show what any assistant says about your brand, but it is a quick way to review your own site before you set up a search routine. For the questions shoppers bring to a brand they have never bought from, see what shoppers ask before buying from a brand they don't know.
Where Hudson South's four steps come in
Our work runs in a fixed order, starting with Authority Asset, which collects the brand's approved knowledge and evidence: tech packs, ingredient documentation, the approved claims list, policy terms, and the answers customer service gives every day. Hudson Twenty-One documents the brand's way of writing and speaking across 21 dimensions, so a fit note or a returns page sounds like the brand and not a template. Market Intelligence researches the questions shoppers ask in your category and what competing product pages already answer. AEO Creation then writes answer-first product content from those inputs. Those steps shape what your pages say; whether an assistant draws on them remains the assistant's choice.
Put one product page next to competitors'
Bring one product and the question shoppers ask about it most. On the free 30-minute competitor analysis call, we compare how your page and competing brands' pages answer it.
Get your free competitor analysisSources
- Shopping with ChatGPT Search. OpenAI Help Center. Read 2026-10-16.
- Publishers and Developers FAQ. OpenAI Help Center. Read 2026-10-16.
- AI Features and Your Website. Google Search Central. Read 2026-10-16.
- How To Add Merchant Listing Structured Data. Google Search Central. Read 2026-10-16.