We tested how AI systems recommend protein powder brands across a set of buyer-style queries. The goal was not to evaluate product quality, but to observe which brands are included in responses — and how often.
Across the queries we ran, one brand appeared in most responses, while others appeared only occasionally or not at all. This suggests that AI visibility is not evenly distributed — even within a well-known category.
In this small sample, ChatGPT treated broad protein powder queries as a search for safe, consensus-backed category leaders, then reshuffled the recommendation set when the prompt introduced a clear constraint such as "without artificial sweeteners."
The pattern was not "best product wins." It was closer to: general queries favour brands that are easy to summarise as trusted all-around defaults, while constrained queries favour brands with an unmistakable association to that need. Brands that were relevant but less clearly anchored to either role were often excluded.
Five "best" protein powder queries were run in ChatGPT in a controlled setup. This was a directional experiment designed to observe recommendation behaviour, not a definitive market-wide ranking study.
Single ChatGPT session, no added brand preferences, no external search workflow, no follow-up prompting. The focus was the model's first-pass recommendation logic.
Identify which brands function as default anchors, which enter only under constraints, and what this suggests about AI visibility in supplements.
Because this experiment used only five queries on one AI surface, the findings should be read as directional evidence of selection logic, not a universal rule for the entire category.
Across five queries, brand appearance was uneven. A small cluster of brands repeatedly surfaced, while many category-relevant brands did not appear at all.
In this sample, Optimum Nutrition behaved like the default category anchor: it stayed present across most broad queries, but disappeared when the prompt demanded a very specific ingredient constraint.
The matrix below shows the shift from broad-query defaults to constraint-specific replacements.
| Brand | Best Overall | Muscle Gain | For Women | No Sweeteners | Best Brand |
|---|---|---|---|---|---|
| Optimum Nutrition | |||||
| PEScience Select | |||||
| Myprotein Impact Whey | |||||
| Naked Whey | |||||
| Allmax | |||||
| MuscleTech Nitro Tech |
The most important shift happened in "without artificial sweeteners": the dominant anchor brand disappeared, and the recommendation set was rebuilt around cleaner-label associations instead.
Three patterns appeared consistently in this experiment. These are not optimization tactics; they are observed selection behaviours.
Pattern 01
In broad "best" queries, ChatGPT repeatedly surfaced brands that can be summarised as safe, familiar, and general-purpose. Optimum Nutrition led this group. The model appeared to reward brands with strong consensus framing — "classic," "industry standard," "widely recommended" — rather than the most specialised or differentiated formulation.
Observed anchor behaviour: Optimum Nutrition appeared in 4 of 5 queries and remained visible across broad recommendation contexts.
Pattern 02
When the prompt introduced a clear filter, broad-query winners lost their advantage. The "without artificial sweeteners" query did not merely reshuffle the ranking; it changed the eligibility criteria. Brands such as Naked Whey, Allmax, Isopure Zero Carb, and LeanFit became relevant because their positioning mapped more directly to the constraint.
Optimum Nutrition appeared in 4 of 5 total queries, but not in the no-sweeteners query. Constraint clarity overrode default anchor status.
Pattern 03
Some brands appeared only when the query strongly matched their positioning. Performance-oriented or engineered formulas showed up in muscle-gain contexts, but did not carry into general "best" queries. In other words, specialisation can improve depth in a narrow lane while reducing visibility across the broader recommendation surface.
MuscleTech Nitro Tech appeared only in muscle gain. Other more specialised products were contextually useful, but not broadly portable across the query set.
This experiment points to a simple three-layer selection model. Visibility was conditional, not absolute.
AI looks for a brand it can safely summarise as a trusted all-purpose answer.
AI applies a more explicit eligibility filter and rewards brands with a clear association to the requested attribute.
AI includes the brand when the query aligns tightly with its role, but not necessarily outside that role.
For brands, this means recommendation visibility is often a positioning problem before it is a product problem.
If you want to appear in generic "best protein powder" prompts, you need more than a good formula. You need to be legible as a dependable, all-around choice that AI can summarise with confidence.
For constrained queries, vague brand messaging is costly. If "clean label," "no artificial sweeteners," or another attribute is central to your brand, that association has to be obvious and consistent enough for the model to retrieve it.
Highly specific products may win exactly where they should, but disappear elsewhere. That is not necessarily a weakness — but brands should understand the tradeoff between broad default visibility and narrow contextual strength.
Several familiar protein brands did not appear in this sample at all, including Ghost, Transparent Labs, Orgain, Garden of Life, and Vega. Their absence does not mean they are weak products; it suggests they were not the model's easiest fit for the specific prompts tested.
In this experiment, ChatGPT did not appear to choose the "best" protein powder in any objective sense. It chose brands that were easiest to defend for the context of the query — broad anchors for general prompts, explicit-fit brands for constrained prompts, and specialised brands only when the framing clearly called for them.
That makes AI visibility less about raw brand awareness alone and more about whether your brand occupies a clear, repeatable role in the model's internal summary of the category.
Every category has its own recommendation logic. We run controlled experiments to show which prompts include your brand, which competitors replace you, and what signals appear to drive selection.