A shopper looking for running shoes used to type "best shoes for flat feet" into a search box and work through a page of blue links. A growing number of them now ask an AI assistant instead, and the assistant answers with two or three named products and, increasingly, a way to buy. The page of links never loads. The brands that are not in the answer are not on the shortlist, and there is no second page to climb to.

This is the next move in a shift we described in our look at how search is splitting rather than dying. The question for any business that sells something is no longer only how to rank. It is whether the model will quote you.

Search returned a list and let you compete on it. The answer returns a verdict.

Call it answer engine optimization, or AEO. It is the work of being the source a model cites when it answers, and it is becoming the new version of what SEO used to do. Most stores are not ready for it, and the few that are cannot yet measure whether it is working.

The page is being replaced by the answer

The clearest signal came from OpenAI, which reportedly told staff internally that "chat is dead" and is rebuilding ChatGPT into what it describes as a superapp, bundling agents, coding tools, and partner services like booking and design directly into the assistant. The chatbot that answered questions is becoming the surface where tasks get done, including buying.

That reframes discovery. When the assistant does the shopping, the consumer never sees a results page, a product grid, or an ad unit. They see an answer, and the answer names a small number of products.

Retailers are starting to notice. Trade coverage now puts the question directly, with headlines asking whether your store is ready to be quoted, because being quoted by the assistant is the new equivalent of ranking on the first page. The difference is that page one had 10 slots. An answer has two or three.

Cited, not ranked

SEO was a contest for position on a list the shopper still read. AEO is a contest to be the source inside an answer the shopper mostly trusts on sight.

The mechanics differ in ways that matter. A ranked page rewarded keywords, links, and freshness. An answer rewards being legible to a model and credible to it. That means structured, machine-readable product data, the kind an agent can parse without guessing, and it means third-party credibility, the reviews, comparisons, and mentions on sites the model already treats as trustworthy.

One retail analysis put the second point sharply: third-party credibility is becoming the discovery advantage, because a model deciding what to cite leans on what independent sources say about a product more than on what the brand says about itself. Your own product page is an input. The wider web's verdict on you is the tiebreaker.

There is a harder constraint underneath all of this. An agent cannot recommend what it cannot read. We have written before about the product-data wall that stops AI agents from completing a purchase, and AEO runs straight into it. A catalog locked inside images, scripts, or a checkout flow no model can follow is a catalog that does not exist to the answer.

The stakes are concentration

The reason this matters more than a normal channel shift is concentration. A search page spreads attention across many results. An answer collapses it onto a few.

Amazon has already moved to supply the picks. Its new Agentic Shopping Assistant technology, first applied with Kate Spade New York, is now being offered to other retailers, which means the company that owns the largest store also wants to own the agent that recommends stores. When the recommender and the marketplace are the same party, the question of who gets quoted stops being neutral.

For a brand, the downside is quiet. If you fall out of the answer, you do not get a ranking report telling you why. You get less traffic from a source you cannot see, attributed to a query you never observed, with no obvious lever to pull.

Nobody has the playbook yet

The uncomfortable truth is that AEO is real but unmeasured. The SEO industry spent 20 years building tools that show you where you rank and why. The equivalent for answers barely exists.

You can ask the major assistants a set of questions and check whether your domain shows up in the citations, and a handful of tools are starting to do exactly that. The signal is noisy, though. The same question can produce different sources on different days, the assistants disagree with one another, and none of them explain the choice. The measurement problem is not a detail. It is the reason the discipline is still forming.

That is the opportunity and the risk together. A brand that starts measuring citations now, across the assistants its customers actually use, will have a baseline while everyone else is guessing. The ones that wait will optimize blind.

What to watch

Three things will signal how fast this hardens. The first is whether the assistants expose a structured way for merchants to be read and recommended, the way search engines eventually published guidelines and schemas. The second is whether independent measurement of citations matures into something a marketing team can act on rather than a curiosity. The third is whether the model makers, the marketplaces, and the payment rails converge on a single answer surface, which would make the concentration permanent.

The plumbing for being discovered is being rebuilt around the answer rather than the page. It is happening now, and it favors the brands that are legible to a model and credible to the wider web. The rest will keep optimizing for a results page that fewer and fewer of their customers ever see.

Sources

If the answer names two products and yours is not one of them, who do you ask why, and what do you change?

Charlie Major is a Product Development Manager at Mastercard. The views and opinions expressed in Major Matters are his own and do not represent those of Mastercard.