“AI agents will do your shopping” has been a conference keynote line for two years, and most of what gets said about it is either science fiction or sales copy. That makes it easy to dismiss, which would be a mistake for one practical reason: the groundwork for selling to agents is the same work that improves your Google Shopping performance today. You don’t have to believe any particular timeline to benefit from doing it.
What’s actually changing about how people buy?
A growing share of buying journeys now starts inside an assistant instead of a search box. Someone asks for a decent 350ml ceramic mug under £20 that ships this week, and software goes looking on their behalf: comparing products, checking availability, shortlisting options. The person sees the shortlist, not your category page.
The honest framing is mobile in 2010. Nobody could name the date mobile traffic would cross desktop, but the direction was obvious, the work took time, and the stores that started early spent years collecting customers the laggards couldn’t see. Agent traffic is at that stage: small, growing, and cheap to prepare for if you start before it matters.
What does a shopping agent read?
Not your homepage. Software evaluating products at scale doesn’t render your hero banner, admire the photography, or watch the brand video. It reads structured data: product feeds, schema.org markup, and the catalogs that platforms like Google and Meta have already assembled from merchant feeds.
That last part matters most. The major surfaces agents buy through are being built on the same product data infrastructure that powers Shopping ads today. Your feed was already your listing. Increasingly, it’s also your salesperson.
Why does feed quality decide whether agents can see you?
People forgive gaps. A shopper who can’t find the dimensions in the description scrolls down, checks the photos, or asks. Software doesn’t. An agent filtering on ceramic, 350ml, in stock matches against fields. If your material field is empty, the capacity is buried in a truncated title, and availability says “usually ships soon”, you don’t lose the sale to a better product. You get filtered out before the comparison starts.
Identifiers matter even more here than in classic Shopping. GTINs and brand are how software confirms your item is the same product it has reviews and price history for. To an agent, an item with no identifiers is a stranger making claims.
What does agent-ready mean in practice?
No mystery, just completeness and accuracy on the fields software matches on:
- Identifiers present: GTIN where one exists, brand and MPN where it doesn’t.
- Titles that state facts: brand, product type, key attributes, in that order, instead of promotional copy.
- Attributes filled: color, size, material, product category. The fields you would filter on if you were the buyer.
- Availability and price accurate and fresh, because an agent that gets burned by a stale price learns to skip that seller.
- One clean source of truth serving every channel, so what Google reads and what Meta reads never disagree.
Read the list twice and you’ll notice it’s also the recipe for a healthy Shopping feed. That’s the point. Agent readiness isn’t a separate project. It’s the same feed work with a longer payoff.
What should you do about it this quarter?
Nothing dramatic. Three steps, in order.
- Audit what you have. Find the empty identifier fields, the truncated titles, the availability values only your platform understands. You can’t fix what you haven’t measured.
- Fix with rules, not spreadsheets. A one-off cleanup decays as soon as your catalog changes. A rule between your source feed and your channels applies the fix to every product you add next year too.
- Keep one clean layer. Fix the catalog once and compile per channel from that, rather than patching each channel separately and letting them drift apart.
This is the product SKULayer was built to be: the layer between the feed you already have and everything that reads it. An audit in plain English, rules you can preview before they ship, and one clean version served to Google, Meta, and whatever sells next. If agents are reading in two years, you’re ready. Meanwhile your Shopping feed gets healthier this month, which is a reasonable deal either way.