A two-week study of solidcore revealed how differently AI systems can represent the same brand depending on what they're asked.
I chose solidcore for Ambianceuse's first public case study because it's a brand I personally admire, and because it makes an interesting subject for this kind of research. This was an independent public study; solidcore is not an Ambianceuse client.
Solidcore’s identity is distinctive, its workout format is specific, and its place in the fitness market is well established. Those qualities give us something concrete to measure against the way AI systems describe it.
Across the study, audience positioning and differentiation came through relatively consistently. Other areas, including factual grounding, authority attribution, and recommendation reasoning, were less stable. That contrast tells us more than simply tracking whether the brand appeared in an AI-generated answer.
What the baseline revealed
The two-week baseline showed moderately strong interpretability overall, but the results varied considerably across prompt types.
Prompts about solidcore's audience, differentiation, and brand perception generally produced more coherent responses. The systems could often describe who the workout appeals to and what distinguishes it from adjacent fitness formats. These were meaningful strengths.
The picture became less consistent when prompts required factual grounding, authority attribution, or a clear justification for recommending the brand. Responses could still describe solidcore favorably while offering less precise practical information or a thinner explanation of why someone might choose it over another option.
Consumers ask more demanding questions as they move toward a decision. Understanding what a workout offers is one task. Comparing it with alternatives, checking specific details, and deciding whether to try it are others. The study showed that stronger performance in descriptive contexts did not necessarily carry through to those more demanding questions.
Where the results diverged
The study's signal stability matrix makes that contrast concrete. Category anchoring was broadly stable, and differentiation remained a strength. But positioning and tier framing were mixed, comparison logic varied, and factual stability was unstable.
| What held up | Where the results weakened | Why it matters |
|---|---|---|
| Audience and differentiation. Responses often described who solidcore appeals to and what distinguishes its workout. | Recommendation reasoning. Positive descriptions did not always translate into a specific, well-supported reason to choose solidcore. | A favorable description offers less decision support when the reasoning behind a recommendation is thin. |
| Recognizable workout identity. Responses often identified solidcore's distinctive workout format. | Comparisons and tier framing. Responses varied in how precisely they positioned solidcore against other fitness options. | Consumers comparing workouts need more than broad similarities to understand meaningful differences. |
| Category anchoring. Systems generally placed solidcore within the appropriate fitness category. | Factual grounding and authority. Practical details could be vague or deferred, while founder attribution was inconsistent. | Correctly identifying a brand does not guarantee accuracy when questions become more specific. |
The problem wasn't always omission
Some of the most revealing weaknesses in the solidcore study appeared in answers that included the brand.
In some contexts, responses used an imprecise category description. In others, founder attribution varied, practical information was incomplete, or a response about recommending solidcore sounded favorable while offering limited support for choosing it.
These are not interchangeable problems. An inaccurate attribution is a factual issue. An incomplete explanation may leave a consumer without the information needed to compare options. Favorable language without a substantive recommendation rationale offers little insight into what makes a particular workout distinctive
That distinction is useful for brands assessing their AI visibility. Rather than treating every appearance as a success and every omission as a failure, the study illustrates why individual answers need to be examined for accuracy, specificity, and relevance to the question being asked.
The finding is not that solidcore lacks a clear identity. It is that the identity presented in AI-generated answers was more consistent in some contexts than in others.
What brands should take from this
The findings point to several distinct diagnostic questions. An inaccurate attribution is a factual problem. An imprecise comparison may be a positioning problem. A recommendation that offers little explanation raises a different question: is the answer giving someone enough useful information to make a decision?
For brand teams, the practical starting point is to examine these differences rather than treating AI visibility as a single measure. That means testing questions consumers might actually ask, reviewing the accuracy and usefulness of the responses, and identifying where the brand's public information may need clarification or stronger supporting evidence.
The study cannot establish exactly why a particular AI-generated answer was incomplete or inaccurate. Nor does it tell us that a specific content change will improve future responses. But it can identify patterns worth investigating and provide a baseline for measuring whether those patterns change.
That is the value of measurement before optimization. Rather than responding to every imperfect answer by producing more content, brands can concentrate on the specific weaknesses the evidence reveals, while protecting the areas where their positioning already comes through clearly.
Insights, Strategy and More
Most Recent in: Featured
Most Recent in: Strategy
Most Recent in: AI
Most Recent in: Mechanics
Founder’s Note
