ChatGPT recommends products in our category. how do we get into that list?

By owning a specific answer, not the category. In the store we measured in depth, the brand was named only on the prompts about the one thing it does uniquely, and not once on the generic “best” prompts. Engines assemble shopping shortlists from product records they can read, reviews on platforms they cite, and buying guides that name you for a situation. The audit tells you which of those three you are missing.

we have product schema on every page. is that not enough?

It is necessary and not sufficient. Schema tells the engine what the product is; it does not tell it that the brand is real or that anyone bought it. In our audit a store with clean product markup received “be cautious, no independent validation” on its own brand name because it had seven backlinks, no reviews on the platforms engines read, and a name shared with two other companies. The entity and review layers decide the recommendation.

we sell in fourteen countries. does GEO change per market?

Yes, and the first thing to fix is usually the locales themselves. The store we measured shipped fourteen copies of the same pages, all in dollars, which the engines and Google both treated as duplicates. Once the alternates are real, the prompts differ by market: in that audit the UK demand for the category was twice the US demand, which changed the priority of everything that followed.

how honest is a bench of one audit?

As honest as we can make it, which is why it is stated at the top of the page. One direct-to-consumer audit across three markets, plus the product-schema and merchant checks that run in every one of our 160 audits, plus the published pattern across all of them. We are not claiming a hundred stores. We are claiming a method that is measured, dated and re-run, and we would rather earn the bench in public than pretend to have it.

what will you need from our developer?

Files and template changes, not a rebuild: product and offer records, canonical and locale rules, and the changes that put the price above the video. We hand them over as a forwardable brief with the code included, your developer applies them, and we verify the live result at the next review. Nothing on the list requires replatforming.

Ecommerce SEO agency: GEO for ChatGPT shopping

009 Agency is an ecommerce SEO agency built around Generative Engine Optimization: we make online stores visible, cited and chosen inside AI answers across ChatGPT, Claude, Gemini, Perplexity, Grok and DeepSeek — and we re-measure it every two weeks against a logged baseline. Shopping is where AI answers changed fastest in 2026: a buyer asks for the best product for a situation and receives a shortlist with links, and a store that is not in the shortlist is not in the sale.

We will say the uncomfortable part first. Ecommerce is the thinnest bench in our dataset: one dedicated direct-to-consumer audit measured across three markets, plus the product-schema and merchant checks that run inside every audit we do. What we know about the vertical comes from that, from the 160-audit pattern, and from the engines themselves. If you want a vendor with a hundred Shopify logos, the list we publish names several. If you want the measurement first, read on.

why stores lose in AI answers specifically

The general failure pattern holds everywhere: known but not recommended. In our published dataset 93% of brands were described accurately by AI and only 4.5% were named when a buyer asked who to pick. Ecommerce adds three problems on top of it, all visible in the one store we measured in depth.

the store is named only where it is the category

A premium sports-equipment brand we audited was named on 4 of 24 prompt runs, and all four were the one thing it does that nobody else does: personalised engraving. On “best racket” and even on the brand’s own positioning, “luxury, design-led”, it was 0 of 24. Perplexity did not name it once. The engines recommend the product that owns a specific answer; a store that describes itself in adjectives owns none.

the brand exists, the entity does not

Ask the engines about the brand directly and the answer was “be cautious, no independent validation”: seven backlinks, no reviews on the platforms engines read, and a name shared with a shoe retailer and a watchmaker, both of which ranked above it for its own name. Product schema and a beautiful site do not create an entity. Reviews, listings and a disambiguated name do, and the audit lists which ones matter in your category.

the catalogue is bigger than the store

The same store shipped 533 URLs in its sitemap for 38 real pages: fourteen locales of the same content, all priced in dollars. Its product pages loaded a 13 MB video before the price, and mobile users waited 18 seconds for the largest element. Engines that fetch a product page live get the same experience, and a shopping answer is assembled in seconds. This is the part of ecommerce GEO that is plain engineering, and it is usually the first thing we ship.

what we check on every ecommerce audit

the ecommerce module · runs inside every audit
layer	what we measure
shopping prompts	Twelve prompts a shopper in your category actually asks, run live on two engines at audit and four on retainer, logged verbatim: which stores and products are named, which pages are cited, and where you sit.
product and offer data	Product, Offer, AggregateRating and Organization records on the pages engines fetch; availability and price consistency; feed hygiene for Google Shopping and the merchant surfaces ChatGPT reads.
rendering and locales	What each AI crawler receives from a product page, page weight before the price, and whether locale variants are alternates or duplicates.
entity and reviews	Name collisions, the review platforms cited in your category, and whether the brand resolves to one organisation across profiles.
citable content	Whether the buying guides and comparisons in your category are written by you or by a retailer, a publisher or a competitor — and which ones the engines lift from today.

what the first 90 days look like

The order is the same as everywhere, because it has to be: readable, then verifiable, then quotable.

Audit. The shopping prompts your buyers ask, run live through the engines and logged verbatim, plus crawler and rendering checks per product template. You get a health score, the shortlist engines currently give in your category, and the three most expensive gaps. Complimentary. what is inside →
Fix the catalogue. Product and offer records, one canonical per product, locales as real alternates, product pages that deliver the price before the video. Shipped as files and template changes your developer applies. technical GEO →
Publish what gets quoted. Buying guides and comparisons for the situations shoppers describe, written to be liftable in one paragraph, and one page per specific answer you can own — the engraving page that worked by accident, done on purpose.
Get onto the lists. The review platforms, gift guides and comparison sites the engines actually cite in your category and market — identified from your own panel, not from a generic outreach list.
Re-measure. Same panel, every review, reported as a delta. Series, not screenshots.

proof, and its limits

No ecommerce client is on retainer yet, so there is no ecommerce program result on this page, and we would rather say that than borrow one. The measured result we can point to is from another vertical: a legal client went from 0/5 to 4/5 engines naming it on its category prompt by day 17 (July 2026), with the log published. The mechanics that moved it — one entity, a page per specific answer, placement in the sources the engines cite — are the same ones a store needs, and the audit shows you where yours stand before you spend anything.

one thing we will not do

We will not fake the review layer. Bought reviews, self-published “best of” lists that only feature you, and marked-up ratings with no reviews behind them are read by the engines as promotional, and a promotional corpus produces a warning, not a recommendation. We saw that verdict in the audit above on a brand that had done nothing wrong except have no reviews at all; it is worse when the reviews are there and invented.

ecommerce SEO agency: GEO for ChatGPT shopping