Recommendation Intelligence Research™ · The Vocabulary Test

Turn on search, and AI stops describing your brand. It starts reading the label.

We already know search changes which brands get recommended: 77% of picks flip when browsing is switched on. This time we held the brands and prompts constant and looked only at the words. With search on, the model reaches for ingredients and specifications. With search off, it reaches for impressions. Retrieval pulls facts. Memory pulls a feeling.

50
Prompts, both conditions
21x
Largest word-frequency shift
27%
Noise floor, search on
47%
Noise floor, search off
Where this comes from

We already knew search changes the brands. Does it change the language?

Our browsing study established that turning search on changes 76.9% of recommended brands, far above the 39.1% you would expect from noise alone. That answered which brands the model reaches for. It did not answer what the model says once it has reached for them.

If search genuinely changes what the model is working from, not just which name comes out the other end, that should show up in the vocabulary itself. A model reasoning from a product page it just read should describe things differently than a model reasoning from whatever impression of a brand it already carries in memory.

The test

Same prompts, twice. Only search flipped.

We ran the same 50 prompts through the model twice, once with search available and once without, holding category constant across both runs. Then we counted word frequency in each condition and normalized by total word volume, so a word that just appears more because search-on responses are longer does not get credit it has not earned.

Prompts: 50, identical both conditions
Conditions: search on vs search off
Held constant: category, prompt set
Scoring: normalized word-frequency ratio
Largest shift: monohydrate, 21x
Status: corroborated, not preliminary
What is shown here. The nine words below are the largest ratio shifts we found, not an exhaustive list of every word that moved. We are showing the words with the clearest signal, not claiming these are the only ones that changed.
The finding

With search on, the model talks like it read the label

Every word at the top of the list is a composition or specification term: monohydrate, whey, ashwagandha, chamomile, creatine, and collagen are ingredient names. Leggings and provides are specification language, the kind of word that shows up when a claim needs to be precise rather than pleasant.

Which words shift most when search is turned on
Normalized word-frequency ratio, search on divided by search off, same 50 prompts and categories both ways
word monohydrate 21x provides 20x whey 17x leggings 15x ashwagandha 14x chamomile 5.9x creatine 5.7x collagen 4.9x third party 3.3x All nine are composition, ingredient, or specification words. None are brand-feeling words.
Without search, the model does not use these words nearly as often. It reaches instead for impressions: general, pleasant, unverifiable language about a brand rather than anything specific enough to be checked against a product page.
Why it matters

Retrieval pulls facts. Memory pulls a feeling.

This lines up exactly with what we already knew about how confident the model is in each condition. In the browsing study, the noise floor, how often the model changes its own answer for no reason at all, sits at 47% with search off and drops to 27% with search on. The model is not just picking different words when it can read a product page. It is guessing less.

That is the whole story in one line: when the model has something concrete to read, it reports what it read. When it does not, it reaches into memory for a general impression and states that instead, with the same fluent confidence either way. The vocabulary shift is not decoration. It is a visible trace of a model doing two different things depending on whether retrieval is available.

What this means: the words on your product page are not just for shoppers. When the model can retrieve your page, the language on it becomes the language the model uses to describe you. When it cannot, or does not, the model falls back on whatever generic impression of your category it already carries, and your specific composition, specifications, and claims never enter the answer at all.
Is your product page even readable when AI goes looking?
Get your free AI Commerce Score™ in 10 seconds and see whether your store gives AI something concrete to retrieve, or leaves it guessing from memory.
Get My Free AI Score