The 5 questions an AI assistant asks before recommending your product

When a shopper asks an AI assistant for a recommendation, a short list can take shape before your team sees a visit. These five practical questions help you test whether your catalog gives assistants the evidence to find, compare, and accurately describe your products. Use the answers to remove avoidable gaps and strengthen your chance of making that list.

Here are the five, in plain language.

1. "Can I read this catalog?"

First, the assistant has to understand your products at all. That means complete, consistently named data: title, price, materials, sizes, key attributes, structured so a machine can parse them. A listing with a title and a price is a guess; a listing with the fields that matter is legible. Subtle inconsistencies hurt, too. If some products say "galvanized" and others say "galv." or "zinc-coated," the machine may treat them as different things. If it cannot read you cleanly, nothing else matters.

2. "Can I tell your products apart?"

Next, the assistant has to differentiate. If a whole category shares near-identical, boilerplate copy, there is no basis to explain why one product fits a shopper's need better than another. Differentiation is what turns a coin flip into a recommendation. Catalogs full of interchangeable descriptions give the assistant nothing to choose with, so it chooses the competitor who reads as distinct.

3. "Can I find the right product for a real query?"

Shoppers ask in constraints: "wireless, noise-canceling, under $200, good for travel." This question is whether the right product actually surfaces and the constraint is honored. A system that returns a $260 pair for an "under $200" request has failed, and that failure is invisible to you. It just looks like the assistant recommended someone else.

4. "Can I trust the claims?"

Assistants are increasingly careful about asserting claims they cannot stand behind, because shoppers punish errors and verify what they are told. Yext's 2026 consumer-search research found that even high-trust AI users overwhelmingly take a verification step after a recommendation (Yext, 2026). When the claims in your catalog are internally consistent and backed by evidence, the assistant repeats them with confidence. When they conflict, say a spec sheet that disagrees with the description, it hedges or drops them.

5. "Does the wider internet vouch for you?"

Finally, the assistant looks beyond your catalog: reviews, brand mentions, authoritative coverage. And shoppers verify what they are told. After an AI recommendation, 62% search Google, 58% visit the brand's site, and 52% click the sources the AI cited (Yext, 2026). The story your catalog tells has to survive that second look.

Why "we get mentioned a lot" isn't the answer

It is tempting to treat AI familiarity as proof of readiness. It is not. In a CatalogSignal CEI Benchmark of 100 brands across 10 verticals and four AI providers (more than 100,000 AI shopping queries), mean funnel accuracy was 47.7%, mean hallucination was 9.3%, and mean commercial harm was 10.6%. Some high-familiarity brands still fell into hallucination-risk positions. Getting named is not the same as getting described correctly, and an accurate description gives shoppers a stronger reason to choose you.

Many of the first fixes start with data you already control: product facts, descriptions, terminology, and access. Pair that work with a review of external sources to see where the wider product story agrees or conflicts. A focused catalog slice gives your team a practical place to start within its existing systems.

That is what a Commerce Eligibility Index™ assessment measures: your catalog's answer to all five questions, scored 0 to 100, with the product-level evidence behind each one. Because the assistant is already asking. The only question is whether you know your answers before your competitor knows theirs.

Get your five-pillar baseline. A baseline CEI assessment scores how AI reads your catalog across all five questions, with the evidence to act on it. Get a preliminary catalog check.

Sources