The Therabody analysis: how the massage-gun pioneer performs in AI search

Therabody is the percussive-recovery pioneer, the kind of brand most shoppers in wellness devices can name without thinking. That makes it a useful test, because if a well-known brand struggles to appear when a shopper asks an assistant for a product, the problem is not awareness, it is whether the catalog is readable to the model doing the asking.

Working from real purchase-intent questions, we asked ChatGPT and noted how often Therabody appeared as a recommendation rather than a passing mention.

The result: 7 of 10

We score the way the assistant behaves, not the way the brand markets. Eight or more out of the set is strong, four to seven is mixed, and under four means the brand is mostly invisible for buying questions.

A 70% result is strong, and it tells us Therabody's catalog is mostly legible to an assistant. The brand shows up across category and use-case questions, which means the attributes a shopper names are sitting where the model can read them. The gaps that remain are specific rather than structural.

What the assistant was really testing

The questions span five kinds of intent, and they get harder as they get more specific. There is the broad category question, the use-case question, the comparison question, the attribute-specific question, and the direct brand-recall question. Brands tend to hold up on the first and last and fall apart in the middle, because the middle is where a query depends on use case, body area, and intensity being present as structured data rather than implied in a description.

Why this happens

An assistant does not read a brand's reputation. It reads the catalog, and it can only recommend a product for 'use case' if that attribute exists as a field it can match. When the attribute lives only in marketing copy, the model has nothing to match, so it reaches for whichever competitor wrote the attribute down. That is the whole mechanism, and it is why a smaller brand with a cleaner catalog can out-rank a household name inside an assistant.

What would move the score

The fix is the same regardless of where a brand lands. Map the catalog to the Shopify Standard Product Taxonomy, write use case, body area, and intensity and the rest of the category's attributes to the product metafields, give each product a line of genuine context about what it is for, and make sure the pages render that content server-side so an assistant reads it without running JavaScript. None of it touches the brand or the design; it just moves the truth into the layer the model reads.

You can see your own score in minutes. The free audit reads your catalog and shows the gaps by question.