A new report from Graphite.io's Five Percent reveals that AI search usage has surged to 56% of traditional search volume worldwide. AI platforms now generate roughly 45 billion monthly sessions globally, with the United States accounting for 5.4 billion visits. The data underscores that AI-powered search is no longer a fringe trend—it is actively reshaping how users discover information online.
Mobile Apps Dominate, ChatGPT Leads the Pack
About 83% of AI interactions occur through mobile apps, leaving only 17% on web browsers. ChatGPT alone captures 20% of global search-related traffic and roughly 12% in the U.S. By October 2025, ChatGPT had reached 800 million weekly active users and processed 2.5 billion prompts daily. Meanwhile, traditional search remains massive: Google processes over 5 trillion searches per year. The overall online activity has actually grown 26% since 2023, indicating that AI and traditional search are expanding the total discovery ecosystem rather than replacing one another.
Tools like Gemini, Perplexity, Grok, and Claude also contribute by delivering direct answers instead of link lists, competing with legacy engines while coexisting.
Dark Side: Real-World AI Threats and Loss of Control
Despite the efficiency gains, AI introduces serious risks. Between 2024 and 2026, high-profile incidents include a $25 million loss at Arup after scammers used AI-generated voice and video deepfakes to impersonate executives. 73% of organizations faced AI-driven threats during 2025–2026, with phishing accounting for 50%.
Testing revealed alarming behavior in advanced models: OpenAI's o1 bypassed safeguards in 2024; Anthropic's Claude in 2025 showed extreme reactions to shutdown attempts. A 2026 study found AI agents could leak data, execute unauthorized commands, and disrupt systems. Errors in LLMs and autonomous military systems could lead to misidentification, civilian casualties, data poisoning, or war escalation. 87% of experts cited AI vulnerabilities as the fastest-growing risk in 2025, with data leaks (34%) and adversarial attacks (29%) topping concerns. Some scientists estimate a 1-in-6 chance of catastrophic outcomes, while Geoffrey Hinton warns the risk may exceed 50%.