Indexed vs. chosen. The only AI readiness framework that survives the agent era.
Most AI search work moves a brand along one axis: can a model read your products. Agentic commerce only pays out on the second axis: can a model match your product to a specific person. A two-by-two for locating your brand, and the tools, in five minutes.
Every brand I talk to has some version of the same sentence ready: “We’ve done GEO.” They’ve cleaned up descriptions, added structured data, maybe published an llms.txt file. Then they ask why ChatGPT still recommends a competitor when a shopper describes herself in detail.
The answer is that “AI readiness” is two different properties, and almost everyone is measuring only one of them. This is the framework we use internally to keep them apart. It fits on a napkin.
Two axes
Axis one: can a model read it? This is legibility. Is the product described in language and structure a model can parse, cite, and compare? Clean attributes, consistent schema, complete fields, crawlable pages. Nearly all GEO and AEO tooling lives on this axis, and it is real work.
Axis two: can a model match it to a person? This is answerability. When a shopper says “warm undertone, combination skin, 12 hour wedding in August,” does the product carry the data a model needs to say yes or no, with a reason? Does the shade carry its undertone? Does the formula declare which skin types it suits? Does the garment say what body shapes its cut flatters? Almost no product data does, because product pages were never written to answer questions about the shopper.
Put them together and you get four places a brand can be.
The four quadrants
Invisible (can’t read, can’t match). Thin descriptions, no structure, images doing all the work. The model can’t find you, and if it did, it couldn’t say why you’d be right for anyone. More brands are here than admit it, especially those with large catalogs and small content teams.
Indexed (can read, can’t match). The product is legible. The model knows it exists, knows its attributes, can quote its price. This is where “we’ve done GEO” lands you. You are in the candidate set, alongside everyone else who did the same work. When the shopper’s question is generic, you might get picked. When it describes a person, you’re a coin flip against identical competitors.
Unproven (can match, can’t read). Rare, but it exists. A brand with deep internal knowledge of who each product suits, held in a merchandiser’s head or a fit team’s spreadsheet, that never made it into the data a model can see. The understanding is real. The model has no access to it.
Chosen (can read, can match). The product is legible and it is answerable. The model can find it, and when a shopper describes herself, the model can explain why this product is right for her specifically. This is the only quadrant that pays out when the question is personal, and in fashion and beauty the question is almost always personal.
Where the tools live
Most GEO tools move brands from invisible to indexed. That is the whole product. Better product text, better schema, a visibility dashboard. Useful, and increasingly a commodity, because the platforms are starting to do it for free.
Moving from indexed to chosen requires something different: a model of the shopper, and product understanding deep enough to match against it. That means extracting the traits a photo contains and a feed does not, structuring who each product is for, and holding it in a form a model can reason over. That is the Product-Customer Intelligence Layer, and it is why we don’t describe Veristyle as a GEO tool. GEO gets you to indexed. We’re built for the other move.
Why the second axis is the one that lasts
Here is the uncomfortable part for anyone who has spent the year on legibility. The first axis is self-erasing. Once every brand is legible, legibility stops being a reason to pick anyone. The second axis is not. A brand that knows who its products are for, at the level of detail shoppers use to describe themselves, has something a competitor can’t copy by cleaning up a feed.
When agents start deciding and checking out on the shopper’s behalf, with the brand’s site never loading, the quadrant you are in is the entire funnel. Indexed brands will be in the comparison. Chosen brands will be in the cart.
Locate yourself in five minutes
Pick your best selling product. Ask an AI assistant two questions about it.
First, a product question: “What is the coverage and wear time of [product]?” If the answer is accurate, you’re legible.
Second, a person question: “I have a warm undertone and combination skin and I need something for a 12 hour outdoor event. Would [product] work for me, and why?” If the answer is specific and correct, you’re answerable. If it’s a hedge, a guess, or a competitor, you know which axis to work on.
A free AI visibility audit runs both questions across your catalog and places every product on the grid. Or book a demo and we’ll go through yours together.