GEO Limitations Exposed: Rethinking SEO in the Age of AI

As the digital landscape continues to evolve, the field of search engine optimization (SEO) has faced numerous challenges in keeping pace with the latest developments. One such challenge is the emergence of generative engine optimization (GEO), a discipline that seeks to optimize content for artificial intelligence (AI) models. While SEO has a well-established foundation of metrics and best practices, GEO is still in its infancy, and the infrastructure to measure its effectiveness is far from complete. In this article, we will explore the limitations of using prompt volume as a foundation for GEO strategies and examine what the most effective teams are doing instead.

At its core, SEO relies on a robust framework of metrics and signals that can be tracked and analyzed. From keyword volume to link equity, SEO professionals have access to a wealth of data that informs their content creation and optimization efforts. In contrast, GEO operates in a vastly different environment, where the signals and metrics are still being developed. The most pressing issue is the lack of accurate and reliable data on AI search volume, which is essential for making informed decisions about content optimization.

The Problem with Prompt Volume

Prompt volume is often touted as a key metric for GEO, but it is fundamentally flawed. AI models do not publish query frequency or search volume equivalents, and their responses can vary dramatically even for identical queries. This is due to the probabilistic nature of AI decoding and the contextual features that are opaque to external observers. As a result, the estimated prompt volume data provided by platforms is often modeled and directionally wrong, rather than being based on actual data.

Why Prompt Volume Fails

• Unlike traditional SEO, where millions of people type the same phrase into Google, AI interactions are conversational and variable, making pattern recognition harder with small datasets.
• The non-determinism inherent in AI models means that the same prompt can produce different responses, making consistent and accurate conclusions difficult to draw.
• Research has shown that even identical prompts can yield vastly different brand recommendations, suggesting that point-in-time AI visibility measurements may reflect volatility rather than durable performance signals.

Estimating Prompt Volume

Many tools query AI models via API to simulate user prompts, but this approach has its own set of limitations. Most AI tracking tools rely on API calls rather than mimicking human interface usage, which can lead to inaccurate results. Furthermore, the API-focused nature of querying data means that results are not aligned with what humans actually search for.

• API calls can yield different results than interface results, although the magnitude and implications of these differences require further investigation.
• The month-to-month stability of AI citations is shockingly low, making it challenging to draw reliable conclusions about content performance.

The Study

A landmark study by and Gumshoe.ai in January 2026 shed light on the limitations of prompt volume data. The research involved testing 2,961 prompts across 600 volunteers on ChatGPT, Claude, and Google AI. The findings were striking: there was less than a one in 100 chance of getting the same brand list in any two responses, and less than one in 1,000 chance of the same list in the same order.

As Fishkin bluntly concluded, any tool that gives a ranking position in AI is essentially making it up. This highlights the need for more nuanced and reliable metrics in GEO, rather than relying on flawed estimates.

Profound’s Approach

Profound, a platform that relies on opt-in consumer panels to source its prompt data, applies advanced probabilistic modeling to extrapolate frequency, intent, and sentiment across broader populations. While this approach may seem robust, it has its own limitations, including the potential for skewed samples and the lack of representative coverage.

Limitations of Opt-in Panels

The opt-in nature of Profound’s consumer panels means that the sample may skew toward more tech-savvy, engaged users, rather than a representative cross-section of how the general population actually prompts AI tools.

Conclusion

In conclusion, prompt volume is an unreliable foundation for GEO strategies. The limitations of AI search volume data, combined with the flawed estimates provided by platforms, make it challenging to draw reliable conclusions about content performance. Effective teams are moving beyond prompt volume and exploring more nuanced metrics that can provide a clearer understanding of AI-driven content optimization.

By recognizing the limitations of prompt volume and embracing more robust metrics, we can unlock the true potential of GEO and create content that resonates with AI models and humans alike.

Key Takeaways

• Prompt volume is an unreliable metric for GEO strategies due to the limitations of AI search volume data.
• Effective teams are moving beyond prompt volume and exploring more nuanced metrics that can provide a clearer understanding of AI-driven content optimization.
• The study highlights the need for more robust metrics in GEO, rather than relying on flawed estimates.
• Profound’s approach, while robust, has its own limitations, including the potential for skewed samples and the lack of representative coverage.
• The month-to-month stability of AI citations is shockingly low, making it challenging to draw reliable conclusions about content performance.

Frequently Asked Questions (FAQs)