Google has announced the Universal Commerce Protocol, which will allow shoppers to buy products directly from AI Mode and Gemini rather than clicking through to a store. Meanwhile eighty percent of B2B technology buyers now report trusting generative AI as much as traditional search when researching suppliers.
Both point the same direction. The moment of purchase is drifting away from your website and toward an assistant that reads your data and decides whether to recommend you.
This is not a reason to panic, and most of the advice circulating about it is premature. But there is a small set of unglamorous work that pays off either way, and it is worth doing while it is still cheap.
Your product data becomes the storefront
When a person browses your shop, a thin product description is survivable because photos, layout and reviews carry the rest. When a machine evaluates your product, the structured data is all there is. It cannot infer quality from your typography.
WooCommerce outputs Product schema by default, but the default is thin. What matters for machine readability:
- Price and currency, present and accurate, including sale price.
- Availability, reflecting real stock rather than always in stock.
- GTIN, MPN or brand, which is how a system identifies your product as the same object it has seen elsewhere.
- Shipping and returns as structured data, not only as a policy page. Assistants surface these in comparisons.
- Aggregate rating and reviews, marked up properly.
The last two are where most WooCommerce stores are weakest and where the competitive gap sits.
Attributes are not decoration
Most stores treat product attributes as filter options for the shop page. In a machine reading world they are the product specification.
If someone asks an assistant for a waterproof jacket under two hundred pounds in navy, size large, in stock, the only way your product is a candidate is if waterproof, colour, size and stock are all structured fields rather than sentences buried in a description.
Go through your attribute taxonomy and ask whether it describes the product completely enough for a system that will never look at the photograph.
Descriptions written for a person who already read a summary
Assistants synthesise. What they cannot synthesise is specific, verifiable detail that exists nowhere else: actual dimensions, materials, what it does not do, who it suits badly.
Generic marketing copy is exactly the content a model discards, because it adds nothing to what it already produced. Specific detail is what gets cited.
The boring technical layer
None of the above matters if the machine cannot read the page reliably.
- Product pages must render server side. Content that only appears after JavaScript is a risk.
- Speed matters more, not less. A crawler that times out simply moves on.
- Check robots.txt is not blocking product pages, feeds, or the crawlers you actually want.
- Keep your Merchant Center feed accurate. It remains a primary structured source.
What not to do yet
Do not rebuild your store for a protocol that is not fully live. Do not buy an AI commerce optimisation product from anyone claiming certainty about how selection works, because nobody outside Google has that.
Do the work that is valuable regardless: complete structured data, real attributes, honest specific descriptions, fast server rendered pages, accurate stock. Every one of those improves your normal search performance and your conversion rate today, and positions you if the protocol becomes significant.
That is the test for any AI readiness advice. If it only pays off in a future that may not arrive, it is speculation. Our note on getting cited in AI answers applies the same principle to content, and what changed in WooCommerce 11 covers the platform side.
Frequently asked questions
What is agentic commerce?
It describes purchases completed by or inside an AI assistant rather than on the retailer website. Google Universal Commerce Protocol is one implementation, allowing shoppers to buy directly from AI Mode and Gemini. The practical shift is that your product data, not your storefront design, becomes what is evaluated.
Does WooCommerce output product structured data by default?
It outputs basic Product schema covering name, price and availability. What it does not fully cover by default is shipping, returns, GTIN or MPN identifiers, and complete review markup. Those gaps are typically filled with an SEO plugin or custom output.
How do I make my WooCommerce products visible to AI shopping?
Complete your structured data including shipping and returns, build out real product attributes rather than relying on description text, write specific verifiable detail instead of generic marketing copy, keep stock and pricing accurate, and make sure product pages render server side and load quickly.
Should I rebuild my store for AI shopping now?
No. The protocol is not fully live and nobody outside Google knows how selection will work in practice. Do the work that improves normal search performance and conversion today, which happens to be the same work that positions you if it becomes significant.
Do product attributes affect AI recommendations?
They are likely to matter a great deal, because attributes are the only machine readable specification of what a product is. A constraint like waterproof, navy, size large, in stock can only match your product if each of those is a structured field rather than a sentence in the description.