What 2 Million LLM Sessions Reveal About AI Discovery — What It Means for SEO
Search Engine Land’s Jordan Koene analyzed nearly two million LLM sessions and surfaced a clear lesson for digital marketers: AI discovery is fragmented, industry-specific, and increasingly invisible to traditional measurement. As Koene writes, “ChatGPT commands 84.1% of trackable AI discovery traffic, but it functions primarily as the default tool for broad-market discovery.” That distinction — reach versus intent-driven productivity — is central to how sites must approach visibility in 2026.
Summary of the findings
Koene’s dataset spans nine industries and shows wide divergence in platform adoption. Copilot and Claude showed rapid growth (Copilot 25x, Claude 13x), while Perplexity and Gemini held narrower, more specialized roles. Perplexity, for example, retained meaningful traction in finance because users there demand verifiable, cited answers; Claude became the go-to for deep analysis and long-context reasoning; and Copilot’s surge reflects the power of being embedded where work already happens.
This fragmentation means visibility opportunities are no longer won just by ranking in a single interface. Instead, brands need to earn presence across the networks, sources, and workflows each model trusts — from enterprise productivity apps to licensed finance datasets.
Why this matters for SEO
The shift from search-engine-first discovery to task-embedded AI discovery changes when and how users find answers. Copilot’s integration into Microsoft apps pushes discovery into the moment of execution; Perplexity rewards authoritative, source-linked content; Claude surfaces content that supports deep analysis. Meanwhile, Gemini’s growing “invisible” surface — where users get answers inside Google’s ecosystem and later perform branded searches — reveals a measurement blind spot that can undercount AI’s real influence.
Satya Nadella’s comments from Microsoft’s FY25 earnings underscore this platform-driven behavior: “Our family of Copilot apps has surpassed 100 million monthly active users across commercial and consumer,” he said, highlighting how Copilot is already embedded into workflows where purchasing decisions form.
Actionable steps for site owners
- Map your audience to LLM behavior. Identify where your buyers work and which LLMs they prefer. B2B and enterprise audiences often use Copilot inside Microsoft apps; finance professionals use Perplexity; developers and researchers skew toward Claude.
- Optimize for context and citations. For finance, legal, and other high-stakes verticals, ensure content is traceable with visible sources, structured data where applicable, and clear citations that models like Perplexity can surface.
- Publish analysis-grade content. For audiences that use Claude, produce deep case studies, reproducible methodology, long-form datasets, and documents that invite large-context ingestion (reports, whitepapers, appendices).
- Invest in brand lift and recall. If Gemini and other models are closing the loop invisibly, focus on brand signals — name mentions, branded search, and offline/upper-funnel channels — to capture downstream conversions that AI discovery initiates.
- Rethink attribution. Move beyond last-click. Build time-lagged attribution models, measure branded search lift, and instrument multi-session journeys to capture AI-originated research that converts later.
- Claim presence in model source networks. For Perplexity and other citation-driven systems, partner with or be listed in the data networks and publishers those models trust. Structured feeds, licensed data partnerships, and timely datasets help.
Implications for measurement and long-term strategy
AI discovery is already redefining the conversion funnel. Marketers must accept that some discovery signals will be obfuscated by model design and platform lock-in. The remedy is twofold: strengthen the signals you can control (brand, citations, data partnerships) and adapt measurement to the new multi-platform reality. Expect AI to move discovery into product and productivity surfaces; optimize for the moments where decisions are made, not just where research starts.
Koene’s report concludes with a strategic takeaway that should guide priorities: be present where your users need to be productive, and tailor your content to the reasoning patterns of each model. As he notes, brands “need a multi-platform strategy that aligns with how users expect to be productive at different moments.”
For more detail, read the original Search Engine Land article by Jordan Koene: https://searchengineland.com/2-million-llm-sessions-ai-discovery-468115
.png)
