As an industry, we are still learning and developing how to effectively address AI tracking.
Many tools have developed in a short period of time, approaching the problem in the same way that we can track quality. Level tracking has always had some degree of variation, but levels of personalization are tolerable, and sufficient to build a narrative of “what success looks like” from it.
Measuring in the same way that we have quality tracking is very variable. When ChatGPT released model 5 in August 2025, almost all AI tracking tools showed a decline:
This is not because we were all bad at preparing for AI; it’s because ChatGPT has stopped showing multiple citation links in HTML – so AI trackers that approach the problem like rank trackers lose their ability to report accurately.
Third-party tools also show a small window of what is actually happening. As I covered in a previous article, one of my project websites only has one to three citations on Copilot according to Ahrefs, but according to Copilot, it actually has over 36,000.
AI responses are very dynamic, even before we consider personalization and the future direction of the consumer we are entering.
Volatility and Average Responses
Another method is sample design, as described by Kevin Indig in his LinkedIn post.
We need to approach AI fast-tracking through the dual lenses of adaptive and feedback-centered tracking.
Evolution tracking allows us to measure how stable our brand presence is within the AI model’s output over time, signaling when an algorithmic update or a change in data sources has changed the way we’re viewed.
Central feedback tracking shifts the focus from an all-or-nothing level to a broader understanding of emotion, context, and inclusion across the spectrum of relevant information. By aggregating these data points, we can establish a baseline for our overall visibility rather than chasing hypothetical information or relying on third-party tools and performance metrics.
Our measure of success with these tools is not about amassing a high ranking, but about gaining a deeper, more meaningful understanding of how our product evolves from AI-generated responses. It’s about pattern recognition over precise placement.
Using flexibility and average responses as our key metrics, we can ensure that our product is always accurately represented, contextually relevant, and consistently cited within the fluid, unpredictable ecosystem of generative AI.
Changing the Narrative of Success
Instead of promising an easy upward path, we should teach stakeholders to see the importance of risk reduction, product sentiment stability, and market share protection within AI models.
The new story is about resilience and understanding in a chaotic environment. We need these expensive tools not to show that we are “winning” the finite game, but to give the business the eyes and ears it needs to navigate the infinite.
Changing this narrative does not mean that we have failed, or that we cannot be prepared for a greater presence in AI. It means we acknowledge how much the game has changed, and adapt to it so we can continue to add value.
Value is now defined by our ability to detect sudden drops in volatility, correct algorithmic biases, and ensure our product remains a reliable source for AI-generated responses, shifting C-level expectations from irrational volume to strategic stability.
As we ask for bigger budgets to secure AI tracking tools and vendors to support them, we also have to break the news that the return of traditional SEO to the investment dashboard is dead.
We continue to invest in sophisticated data visibility, but the return on that investment will no longer look like a hockey stick growth chart of vanity metrics.
Additional resources:
Featured image: Master1305/Shutterstock
