Why an AI knowing a brand by name is not evidence that it can discover the brand under competitive conditions.
The system must know what the entity is
Generative systems retrieve information about organizations, products, locations, people, and brands—not only URLs. A technically healthy website can still have an entity-level problem if the system cannot consistently identify what the brand does, its category, location, products, aliases, or relationships.
Recognition and discovery test different capabilities
A branded prompt names the target and primarily tests recognition. An unbranded prompt describes a need or category and tests whether the system independently surfaces the brand among competitors.
“What does Brand X do?” is easier than “Which brands solve this problem?”
The second question is usually more informative for competitive GEO.
Small brands often face an upstream visibility floor
Kumar’s observational study reported a large first-run unbranded visibility gap across its brand cohorts—about 73% for Tier 1, 44% for Tier 2, and 11% for Tier 3—while branded recognition was much higher.
These figures are descriptive rather than causal, but they illustrate why a smaller brand may first need to become a recognized candidate before fine-grained citation optimization becomes the main constraint.
Use a recognition-discovery matrix
High branded recognition with low unbranded discovery means the system knows the entity but rarely selects it competitively. Low recognition and low discovery point toward an entity-foundation problem. High performance on both suggests a strong entity presence.
Questions about this topic
What is a branded prompt?+
It explicitly names the company or product and primarily tests whether the system recognizes the entity.
What is unbranded discovery?+
It measures whether a brand appears for a category, problem, or use case when the brand is not named in the prompt.
Why is recognition insufficient for competitive GEO?+
A system can accurately describe a named brand while never selecting it as a candidate or recommendation for an unbranded need.
Are the 73%, 44%, and 11% figures universal?+
No. They describe cohorts in Kumar’s 2026 observational study and should not be treated as permanent platform benchmarks.
References
Sources are listed in APA 7 style. Preprints are identified as such and should not be treated as peer-reviewed findings unless separately published.
- Kumar, P. (2026). Generative engine optimization at scale: Measuring brand visibility across AI search engines [Preprint]. arXiv. https://doi.org/10.48550/arXiv.2606.20065
- Martinez, O. (2026). Optimizing visibility in generative engines: A critical survey of generative engine optimization (2023–2026) [Preprint]. arXiv. https://doi.org/10.48550/arXiv.2607.14035
- Chen, M., Wang, X., Chen, K., & Koudas, N. (2025). Generative engine optimization: How to dominate AI search [Preprint]. arXiv. https://doi.org/10.48550/arXiv.2509.08919