Getting your WooCommerce products into Google AI Mode shopping answers

Getting a WooCommerce product into a Google AI Mode shopping answer takes two things at the same time: a Merchant Center listing that makes it a candidate, and product content specific enough for a conversational answer to choose it. Winning ordinary Shopping results is a different achievement — Productrise tracked more than 100,000 shopping queries and over two million listings across the US and UK during 21 days in July 2026, and found only about 0.8% of the products visible in standard search also appeared in AI Mode for the same query on the same day (published 29 July 2026, checked August 2026).

That 0.8% changed how we think about catalog work on Contexta. A store can pass every feed check we run and still be missing from the surface it cares about, because feed validity is a candidacy test while AI Mode's pick is a fit test — and fit gets judged on qualities a product feed historically never carried.

Does a Merchant Center feed alone get your products into AI Mode?

No — it makes them eligible and nothing more. The feed is how your product enters the Shopping Graph that AI Mode draws candidates from, so without it you aren't in the running; with it you're one of a very large pool competing for very few slots. The gap between those two states is where most WooCommerce stores are stuck.

Productrise's July 2026 tracking puts numbers on how narrow the surface is. Standard search returned at least one product on roughly 88% of shopping queries against about 23% in AI Mode, and when both surfaces did show products, standard search averaged about 22.5 listings per results page against about 4.3 in AI Mode. Multiply the two effects and you get their headline: roughly 95% fewer product listings overall on the same query set.

Treat that as a shape rather than a scoreboard: Productrise sells organic shopping rank tracking, the window was three weeks, and only two markets were sampled. The finding that survives those caveats is the one that matters here — the two surfaces are drawing largely different subsets from the same graph, so carousel placement is not a proxy for AI Mode presence and can't be used as one. Everything about getting a WooCommerce catalog into Merchant Center in the first place is still necessary; it just stopped being sufficient.

What decides which products AI Mode actually picks?

Fit with the sentence the shopper typed. AI Mode questions are long and loaded with constraints — "a quiet air purifier for a small bedroom that doesn't need proprietary filters" — and your feed can only answer the hard half of that: price, availability, identity, category. The soft half lives in your content, and a product that clears one half without the other doesn't get chosen.

We watch this fail in one direction almost every time. Stores fix identity and offer fields — the work that makes an agent's structured record of the product complete and consistent — then wonder why a conversational answer still skips them. The record is complete and it says nothing about noise, filter type or room size, so when the model reasons over four slots it picks the products whose data speaks to the actual question.

The inverse fails too, and faster. Beautiful product copy on a page with no Merchant Center listing isn't in the candidate pool at all, so nothing gets to read it. Neither half is optional, which is exactly what makes this awkward to staff: the feed belongs to whoever runs the shop, the copy belongs to whoever writes, and the thing being optimised now needs both people in the same file.

What are Google's conversational attributes, and why do they change the job?

They're a set of Merchant Center attributes that carry product content rather than commerce data, and Google states plainly that they exist to "help AI systems and conversational agents better understand your products' specific nuances" (Merchant Center Help, conversational attributes, checked August 2026), naming AI Mode in Search as a destination. Six of them are documented, and most have no equivalent anywhere in a traditional feed.

Read that list as a statement of intent. Google is asking merchants to submit the answers to questions shoppers ask, the documents that prove specifications, and the map of what works with what — the material a conversational answer needs and a price-and-availability feed can't provide. It reframes the feed as a content channel, which is a genuinely new job for a file most stores treat as plumbing.

The honest caveat: these attributes are new as of mid-2026 and there is no public performance data showing what submitting them does to visibility. What can be said with confidence is that the same data underpins the entity your listing attaches to, and everything the Shopping Graph does with your feed starts from that match — so richer, non-contradictory product data has never been the wrong direction, even before anyone can quantify this specific bet.

How do you get that data out of WooCommerce?

You create it, because none of it exists natively. WooCommerce has no field for a product Q&A, no field for a linked spec sheet, no structured way to say "this filter fits that purifier", and no popularity rank — so unlike GTIN or brand, these aren't fields sitting empty in your catalog waiting to be filled. They have to be authored and then mapped, in practice through a supplemental feed or custom product attributes your feed tool can read.

Three failure modes are worth knowing before you start. Q&As written as marketing rather than answers are useless, because the value is in the specific fact ("the 2019 model needs adapter B") not the enthusiasm. Related-product entries pointing at identifiers you never populated resolve to nothing. And any answer you submit that contradicts your product page gives the system a reason to distrust the whole record — the same consistency problem that sinks price and availability, now extended to prose.

The writing is the part that stalls, since a 400-SKU catalog means 400 sets of answers nobody has time to draft. On our side that's the job Contexta's AI editor does: it generates answer-first copy in the site's own learned voice against the page's real Search Console queries, pulling prices and specs from the live catalog so it states what you actually sell instead of inventing it. It writes the content — it doesn't build your Merchant Center feed, so the mapping into question_and_answer and the rest is still yours to wire up.

How do you tell whether any of it worked?

Through Merchant Center's AI performance insights, and only partially. Google announced the report on 27 May 2026 (Merchant Center Help, checked August 2026), covering AI Mode, AI Overviews in Search and the Gemini app, with share of voice against competitors as the headline metric alongside funnel performance, product term insights and a view of missing product attributes — rolling out to the US, Canada, Australia, India and New Zealand.

Share of voice is a relative measure, not a click count, and that limit is the whole story of measurement on this surface right now. Search Console reports AI Mode impressions but no click data, so between the two tools you can see whether you're being shown and roughly how you compare, and you cannot yet trace a sale back to an AI Mode answer — a gap worth keeping in view alongside how AI Mode differs from AI Overviews as a surface.

That argues for a specific order of work rather than a catalog-wide rewrite. Confirm your products are actually in the graph, because nothing else counts until they are. Then take your twenty best-selling SKUs, write real answers to the questions your support inbox already receives, map them, and watch share of voice on those product terms rather than across the catalog. With roughly four slots in play on a fraction of queries, this surface will have fewer winners than the carousel did — and as of August 2026 anyone promising you a reliable route into those four slots is describing a system whose selection logic nobody outside Google can see.