A
argbe.tech - news1min read
2 million sessions show AI discovery is splitting by workflow, not one “winner”
An analysis of nearly 2 million LLM discovery sessions suggests ChatGPT remains the default, while Copilot, Claude, and Perplexity win specific job-shaped moments. The data points to platform-specific optimization rather than a single “AI search” playbook.
Search Engine Land analyzed nearly 2 million LLM discovery sessions across nine industries (spanning 2025) and found that “AI discovery” behaves differently depending on the job users are trying to finish.
- ChatGPT accounted for 84.1% of trackable AI discovery traffic in the dataset, acting as the default starting point in many categories.
- Growth rates diverged sharply across major tools: Copilot grew 25×, Claude 13×, and ChatGPT 3×, while Gemini stayed roughly flat and Perplexity rose only ~1.15× overall.
- In SaaS, education, and finance, Copilot’s growth outpaced ChatGPT by wide margins (21× vs. 2×; 27× vs. 6×; 23× vs. 4.2×, respectively).
- Perplexity’s sustained foothold showed up mainly in finance, where it held 24% share; in other verticals its share fell steeply (e.g., publishers 41.5% → 3.6%, education 28.5% → 5.2%).
- Claude represented 0.6% of total discovery traffic, but surged in research-heavy segments (publishers 49× growth; finance 38×; education 25×).