Edited on Jun 18, 2026.
The 2011 thesis that Facebook would overtake Google as the dominant search engine turned out to be directionally correct about the structural threat — and wrong about the source of it. Google's search dominance did get challenged. It did not get challenged by Facebook. It got challenged by ChatGPT, Claude, Perplexity, Gemini, and Google AI Overviews — the answer-engine layer that did not exist in 2011 and that now mediates an estimated 30%+ of US consumer information-seeking behavior under 35.
What 2011 got right
- The semantic-search thesis. The next dominant search experience would understand intent, context, and personalized synthesis rather than keyword-and-link matching. Correct. The current answer-engine paradigm runs on exactly this.
- The talent-wars observation. OpenAI's 2015–2024 hiring spree pulled senior Google AI researchers. Anthropic's founding team was largely former OpenAI safety leadership. Perplexity, xAI, and Meta AI all hired aggressively against Google's research ranks.
- The data-as-competitive-advantage thesis. Whichever platform had the deepest training corpus could build the most personalized search experience. Correct.
What 2011 got wrong
- The assumption that Facebook would be the challenger. Facebook never built a meaningful web-search product. The 2013 Graph Search launch was a within-Facebook search feature, not a web-search product. Facebook's data moat — social-graph behavioral data — was the wrong kind of data for the next search paradigm.
- The assumption that switching costs in search were trivial. Apple's Safari default search deal with Google (worth ~$20 billion annually, subject to the 2024 DOJ antitrust ruling) is the structural moat.
- The underestimation of Google's ability to evolve. Google AI Overviews (2024), Bard (2023), Gemini (2024).
Where the search competition actually went
- Google still dominates traditional search. Approximately 89% global market share (mid-2026), compressed from 92% in 2020.
- The answer-engine layer is the new competition. ChatGPT ~700M weekly active users. Perplexity hundreds of millions of monthly queries.
- The Apple search-defaults deal remains the structural moat. $20 billion annually, subject to the 2024 DOJ ruling.
- Meta did not build a search engine. Meta built LLaMA, Llama 2/3/4, and AI assistants integrated across Instagram, WhatsApp, Facebook, Messenger.
- 30%+ of US consumers under 35 begin product research with an AI engine rather than Google.
The AI Communications implication
Brand visibility in 2026 is no longer measured by Google search rankings alone. It is measured by Citation Share — the share of answers a brand earns when ChatGPT, Claude, Perplexity, Gemini, and Google AI Overviews respond to category-relevant prompts. The discipline is called Generative Engine Optimization (GEO). The category is called AI Communications.
