ChatGPT Search and Perplexity are the two AI search engines generating the most direct referral traffic right now, making them priority targets for GEO optimization. ChatGPT Search launched in late 2024 and already shows up in analytics for sites with strong expertise signals and citation-worthy content. Perplexity has been actively crawling and citing sources since 2023, and we see consistent referrals from it across our portfolio, especially for how-to content and comparison queries. Google AI Overviews technically reaches more users because it's integrated into standard Google Search results, but the traffic pattern is different. AI Overviews appear for roughly 15-20% of queries in most industries, but they often answer the question directly without sending clicks. When they do cite sources, it's typically toauthoritative domains with structured data and clear entity relationships. Microsoft Copilot and Claude with search integration are growing but still represent smaller traffic volumes. Copilot pulls heavily from Bing's index plus real-time web results, while Claude's search capability is newer and less commonly used for discovery queries compared to research tasks. For actual GEO optimization work at Ottawa SEO, we focus on: - Structured content with clear claims and supporting evidence that AI models can extract and cite - Entity optimization through schema markup and consistent NAP data for local queries - Concise, quotable statements in the first 100 words of key pages - Lists, comparisons, and data tables that AI engines parse cleanly - Author credentials and E-E-A-T signals that increase citation likelihood The landscape shifts monthly. We track referral sources in GA4 and server logs to see which AI engines actually send traffic versus which just scrape content. For most businesses, optimizing for traditional Google Search and AI Overviews simultaneously makes sense, then layering in Perplexity and ChatGPT-specific signals as those platforms mature. The citation formats and ranking factors aren't fully documented by any platform, so testing and measuring actual referral patterns matters more than speculation.