January 30, 2026
8 min read

Case Study - Travel Agency

by Whitebox Team

In our previous article, we showed that AI search isn't unpredictable - it's simply not being measured correctly. When LLM source usage is tracked over time, clear patterns emerge, with certain sources steadily gaining weight inside ChatGPT responses, especially in the U.S. market.

But identifying a signal is only the first step. The more important question is what happens when that signal is turned into action.

In this follow-up, we share a real Travel Agency brand case where measured trust signals were translated into focused GEO strategy - and produced measurable improvements in LLM visibility and sentiment.

Turning LLM Trust into Measurable Impact

Measurement alone is not the goal. A signal only becomes meaningful when it drives a decision - and that decision produces a measurable result. This second phase shows what happens when LLM trust signals are not just observed, but deliberately translated into action.

At the end of Phase One, we showed that Trustpilot was steadily gaining relative weight in ChatGPT responses across multiple U.S. focused entities. One of those entities was a Travel agency brand operating in a highly competitive category, where trust and credibility strongly influence user decisions.

In November, Trustpilot was already present in LLM outputs for the brand, but with relatively low weight (7.4%), indicating early-stage influence rather than established dominance. At the same time, the sentiment associated with that presence was strongly negative.

With a low sentiment score of -0.74 on a scale from -1 to +1, the content reflected predominantly negative user experiences. As a result, the model was learning both the growing importance of the source and the negative sentiment attached to it, making sentiment the variable that needed to be addressed through deliberate action.

By December, its relative weight had increased substantially, signaling growing model reliance. This trend, rather than a single snapshot, drove the decision to focus action on this source.

November-December baseline graph showing Trustpilot weight and sentiment line

From Signal to Strategy: GEO in Action, Travel Agency Study

The insight from Phase one was not simply that Trustpilot mattered, but where and how strongly it mattered. The signal was clearest in U.S. based outputs, reinforcing a key principle we consistently observe: AI search behavior is local, not global.

Based on that signal, Whitebox recommended a focused GEO action item that treated Trustpilot as a priority trust surface for ChatGPT, emphasizing active engagement to strengthen sentiment.

This included encouraging positive user feedback, responding directly to complaints to resolve issues, and reinforcing favorable reviews to mitigate negative signals, all while concentrating efforts on U.S. specific visibility rather than applying a broad, global approach.

This was not a broad reputation campaign. It was a precise intervention, aligned directly with how LLMs were already learning.

Elevate Negative Sentiment - Whitebox platform interface showing Trustpilot source weight over time

The Impact

By January, the impact of that decision was clearly reflected in LLM outputs. Trustpilot's relative presence increased, but more notably, its sentiment shifted decisively upward, an 80% improvement in positive sentiment.

This transformation was not immediate, nor was it driven by a single spike in activity. Instead, it unfolded gradually, following the same steady, reinforcing the pattern observed in Phase One.

Sentiment Score and Weight chart showing improvement from November to February

This matters because LLMs do not simply count mentions. They internalize patterns: which sources appear repeatedly, which carry stable sentiment, and which remain useful across contexts.

Closing the Loop

This case closes the loop that Phase One opened. Measurements revealed a growing trust signal, Cross-industry analysis across multiple countries revealed that this signal was geographically concentrated and actionable insights focused engagement on the platform LLMs were already learning to trust.

The outcome was a sustained improvement in GEO visibility and sentiment.

This is what GEO performance looks like when it is driven by data and tailored actions rather than assumptions, instead of reacting to AI search as unpredictable.

Whitebox customers turn AI trust signals into a clear, actionable growth engine.