AI shopping agents are a storefront problem, not a marketing one
A small share of traffic today, a large one soon. The work is making the store legible to a machine that wants to buy on someone's behalf.
When Outlook Business asked me about AI shopping traffic last year, my honest answer was: it's probably 1% or less of the purchase funnel right now, and within a year it could be 10–15%. At that scale it is a significant amount of money, and the only question that matters is whether the customer can actually complete the purchase.
What an agent needs
An agent does not scroll. It reads. So the things that used to be SEO hygiene become the product:
- Structured data that is complete and true — price, availability, variants, shipping, returns.
- Product pages that state the facts in text, not in a banner image.
- A checkout that a machine can get through without a captcha, a modal, or a surprise.
- Policies written for a reader that will quote them back to your customer.
What it changes
The storefront stops being a page and becomes an API with a human-readable view. That is not a marketing project. It is a technology project that marketing benefits from.
We have been doing the boring version of this for a while: making sure the catalogue's SEO fields exist, that schema is present, that the checkout is clean. The agent era just raises the stakes on work that was always worth doing.