Search used to start with a few keywords: "hiking jacket waterproof". Now a shopper types or says the whole situation: "a waterproof jacket for hiking in Scotland in spring, light enough to pack, under €200". Google's AI Mode, ChatGPT, Perplexity or Gemini answers with a shortlist and reasons, often before the shopper visits a single store.
For a brand that sells through its own store, the question is no longer only "do we rank?" It's "are we on that shortlist, and does the answer explain why?"
Three ways a store gets into the answer
How AI search reaches a D2C store
| What happens | What the store needs | |
|---|---|---|
| 1. Cited source | AI quotes a passage from a product, category, technology or guide page and links to it | Indexed pages, key content in the HTML, answers stated clearly near the top |
| 2. Product card | AI shopping results show the product with price, availability and rating | Complete product feeds, valid Product markup, attributes that match the page |
| 3. Recommendation | The product is named as the pick, with reasons: who it's for, what it does better | Store pages that state those reasons in plain text |
The third one matters most and is the one stores most often lose. A model takes the reasons for a recommendation from whichever source states them most clearly. If the store page only names a technology and shows a size chart image, the reasons get borrowed from a retailer or a reviewer, in their words, sometimes wrong.
One store, every engine. There is no separate content for Google and for ChatGPT. All three routes draw on the same clear, consistent product information.
What an AI-ready store looks like, page by page
Most stores already have the foundation: titles, specs, filters, reviews. What's usually missing is a second layer that states answers and reasons in text.
Product page
- Opening lines that say what it is, for which activity, conditions or need
- Fit, sizing and compatibility answered in text, not only in a chart image
- Each technology or ingredient explained in one plain sentence, not just named
- A short FAQ built from real shopper questions
Category page
- A short intro that answers the category question, such as "which jacket for wet, cold hikes?"
- Filters and labels that match how people ask: activity, conditions, fit, budget
- A "which one is right for me?" block or comparison
Technology and collection pages
- What the technology does, with measurable facts where they exist
- Which products use it and how they differ
- Comparisons across the range as HTML tables
Guides and help content
- Size, fit, care and "how to choose" guides as live text, not PDFs
- Headings phrased as the questions people ask
- Steps and tables in HTML, not in images
Site-wide, the basics still decide everything: content in indexable HTML, AI crawlers allowed, structured data that matches the visible page, and the same product facts on the store, at retailers and in feeds.
Images, video and comparison charts
Visuals still win the shopper's attention and they should stay. AI assistants can read text inside images, but it costs them more than reading plain text, so today it tends to come lower in their priorities.
Visual assets · seen by shoppers, read by AI
Do
- Repeat key claims from images in the page text
- Build comparison charts as real tables
- Give videos descriptive titles and repeat their claims in copy
- Fill in alt text and every product attribute
Don't
- Keep size charts only as images
- Put the key benefit only on a banner
- Hide guides in PDFs
- Let the feed say something the page doesn't
No one knows how much weight assistants will give to visuals next year. One set of claims carried across copy, images and feeds works today and holds either way.
Fix templates, not pages
A store with five hundred products doesn't need five hundred rewrites first. It needs each page type fixed once: which blocks it has, which facts go where, which structured data it carries. Every product built on that template inherits the improvement, including the ones added next month.
Retrofitting pages one by one costs far more than fixing a page type once. It's also where most store projects stall: good copy written into a CMS that has no field for it, or a FAQ that exists in a document but never makes it onto the template.
How I approach a store
- Baseline. Where the brand appears in AI answers today versus named competitors, plus a technical and content check of the store.
- Question map. Real shopper questions from search data, AI prompts, site search, reviews and customer service, mapped to page types.
- Template guidelines. One set of rules per page type: which blocks, which facts, which structured data.
- Content at scale. AI-assisted drafts from approved product data, reviewed by people and localised per market.
- Build it in. Answer components and structured data added to templates; feeds aligned with what the page says.
- Measure. AI visibility, citations, AI referral traffic and search performance, in Search Console, GA4 and AI tracking.
The principle across all six: optimisation, not reinvention. Nothing is invented; the store says what the brand already knows about its products, in a form a model can use.
What a store can't fix alone
Honest limits matter here. A store controls its own pages and feeds. It doesn't control what retailers, reviewers and media write, and models check whether others agree before they recommend. If the store says one thing and every retailer says another, the store rarely wins.
That's why store work goes together with consistent product data at retailers and some Digital PR. If you run a D2C store and want to know whether AI recommends it, or borrows your reasons from someone else, that's where I'd start: an e-commerce SEO audit with an AI-visibility baseline.
Based on AI-search work for consumer brands selling through their own stores and through retailers; examples are generic.
FAQ
Frequently asked questions
In three ways: as a cited source when it quotes and links a store page, as a product card in AI shopping results built from feeds and structured data, and as a recommendation when it names the product as the pick with reasons.
Opening lines that say what the product is and who it is for, fit and compatibility answered in text, technologies explained in plain sentences, a short FAQ from real shopper questions, and structured data that matches the visible page.
It can, but it costs more than reading plain text, so it tends to come lower in priority. Repeat the key claims from images, charts and videos in the page text to be sure they are picked up.
Yes. AI shopping results with price, availability and rating are built largely from product feeds and structured data. Feeds should be complete and say the same as the page.
Start with templates instead. Fixing each page type once, with the right blocks, facts and structured data, improves every product built on it and costs far less than rewriting pages one by one.
No. Google AI features, ChatGPT, Perplexity and Gemini draw on the same clear, consistent product information. One set of facts, well structured, serves all of them.
