A customer asks an assistant to find them waterproof boots under two hundred dollars, in stock, delivered by Friday. The assistant comes back with three options. Your store either appears in that list or it doesn't, and nobody tells you which.
That's a different game from search. An agent doesn't browse, it parses. It reads your data, checks your availability and tries to complete a purchase, and it gives up quietly the moment any of that gets awkward.
Agents read data, not design
The pretty parts of your product page are invisible to an agent. What it can use is the structured layer underneath: Product and Offer markup, price, availability, dimensions, materials, return window, shipping time. If those live only in a tab, a modal or an image, they may as well not exist.
The same goes for your feed. A clean product feed gives an agent a reliable source without crawling every page, and a feed that lags behind real stock is worse than no feed at all, because it produces confident recommendations for things you can't ship.
What makes you buyable is dull: standard checkout fields labelled properly, with payment and shipping options visible. An agent filling a form is doing what a screen reader does, and the same discipline serves both.
What makes you skippable is also dull: custom checkout scripts, policies hidden behind modals, ambiguous availability, prices that only appear after a click. None of it looks broken to a person.
The uncomfortable part is that none of this is new work. It's the work that was already on the list.
Structured data was an SEO chore, and it's now the interface. Accurate stock was an operations problem, and it's now a discovery problem, because an agent that recommends an out of stock item learns not to trust your store. Fast, server rendered pages were a conversion tactic, and they're now table stakes for anything parsing your site against a timeout. Clear policies were a support cost, and they're now selection criteria. Every one of those was already worth doing for humans, which is why the stores that do well with agents will mostly be the ones already doing well on fundamentals.
What is new is the tolerance for ambiguity, and there isn't any. A shopper who can't tell whether shipping is free will guess, ask or take the risk. An agent comparing three stores just picks one of the other two.
An agent that can't read your stock will not recommend your store
Structure is the new storefront
Which is why this lands as engineering rather than marketing: schema and feeds on Shopify or Magento, content modelled properly in Drupal, and the integration that keeps availability honest.
You don't need an agent strategy. You need product data an agent can read, stock it can trust and a checkout it can complete, which is the same list your best customers have wanted all along.
If you want to know where you stand, ask whether your product page can answer price, availability, shipping and returns without a human clicking anything. That's where we start too, across APIs and integrations and the stack underneath your store. And if your stock numbers are the weak link, that's its own problem.