2 Million LLM Sessions: Key AI Discovery Insights
▼ Summary
– ChatGPT dominates AI discovery traffic (84.1%) but functions primarily as a default tool for broad-market discovery, necessitating a multi-platform strategy.
– Different LLMs are winning in specific industries: Copilot grows fastest in B2B verticals like SaaS and finance by integrating into Microsoft workflows, while Claude excels with developers and publishers for deep, standalone analysis.
– Perplexity maintains a significant 24% market share only in finance, where its cited, verifiable answers meet the industry’s need for trust and auditable data.
– Gemini’s declining tracked traffic likely signals an attribution collapse, not a user decline, as it keeps users within its interface, making AI-driven discovery harder to measure.
– Success depends on aligning with platform-specific user behaviors: optimize for Copilot in enterprise workflows, Perplexity for finance citations, Claude for technical analysis, and monitor branded search to account for fragmented, multi-session journeys.
An analysis of nearly two million large language model sessions across nine industries throughout 2025 reveals a fragmented and nuanced landscape for AI discovery. The initial assumption that ChatGPT dominates with uniform usage patterns is incorrect. While it commands over 84% of trackable traffic, it primarily functions as a default tool for broad discovery. The strategic imperative has shifted; brands must now adopt a multi-platform strategy aligned with how users achieve productivity in different contexts. Success hinges on identifying which platforms actively enable user workflows versus those that merely support initial exploration.
The data shows a significant divergence in growth rates among the major players. ChatGPT experienced a 3x increase, but Copilot and Claude grew at rates eight to ten times faster, at 25x and 13x respectively. In contrast, Perplexity and Gemini showed minimal growth, effectively reinforcing their roles within specific, tightly defined knowledge workflows. These aggregate figures point to deeper strategic priorities, as highlighted by industry leaders focusing on platforms that deliver tangible user value within particular ecosystems.
Copilot’s explosive 25x growth is most pronounced in B2B verticals where professionals are already embedded in the Microsoft ecosystem. In SaaS, education, and finance, Copilot adoption skyrocketed because it integrates directly into the tools, like Excel and Teams, where work already happens. The key insight is that discovery is moving into the workflow itself. Visibility is no longer won during preliminary research but during execution, when user intent is at its peak and decisions are actively being formed.
For Perplexity, the story is one of stark specialization. Its overall growth was flat, but it maintains a commanding 24% market share in the finance industry. This is the only sector where it holds sustained, meaningful traffic. The reason is foundational: financial decisions demand verifiable citations. Perplexity’s partnerships with data providers like FactSet and Morningstar, along with features that provide traceable sources, meet the non-negotiable need for trust and auditability in finance. In most categories, convenience wins, but here, verifiability is paramount.
Although Claude represents a small fraction of total traffic at 0.6%, its concentration is highly revealing. It dominates in fields requiring deep, standalone analysis, publishing, education, finance, and SaaS, where professionals use its extensive context window for strategic synthesis. Unlike Copilot, which aids task execution within a workflow, Claude acts as a reasoning partner for complex analysis, such as evaluating lengthy manuscripts or legacy codebases. The audience is smaller but wields significant influence, requiring content built for deep analysis rather than superficial summaries.
The situation with Gemini presents a potential measurement crisis. Its tracked traffic shows confusing patterns, likely indicating an attribution collapse rather than a user decline. Gemini increasingly keeps users within its interface, delivering synthesized answers without prominent source links. This makes AI-driven discovery invisible to analytics, as users may later convert via branded search. This suggests that the true volume of AI-assisted research is likely significantly higher than what current tracking can capture, breaking traditional last-click attribution models.
Choosing the right LLM strategy depends entirely on your audience. For enterprise audiences working within Microsoft tools, Copilot is essential. For high-stakes financial decisions, visibility within Perplexity’s network of verified sources is critical. When targeting technical evaluators and strategists, optimizing for Claude’s deep-analysis capabilities is key. In emerging categories, starting with ChatGPT’s broad reach is wise, but be prepared for platform migration as the audience matures. Across the board, marketers must build measurement models that account for multi-session, cross-platform journeys and the growing gap in attribution.
The future of AI discovery is not about a single platform. It’s a fragmented landscape where understanding your audience’s specific workflow, need for verification, and depth of analysis determines which LLM becomes their tool of choice.
(Source: Search Engine Land)
