How people actually use ChatGPT vs Claude - and what the differences tell us (www.zdnet.com)

OpenAI and Anthropic released large-scale usage analyses that together map how people actually interact with consumer and enterprise LLMs. OpenAI’s study of millions of ChatGPT conversations (May 2024–June 2025) found non‑work use overtook work use (work messages fell from ~47% to 27%) even as daily messages surged from ~451M to 2.6B. Top uses are practical guidance, information-seeking, and writing (≈80% of chats); coding is only 4.2%. OpenAI groups interactions as Asking (49%), Doing (40%) and Expressing (11%), and notes that when ChatGPT is used for work it’s often for “doing” tasks like editing—over 40% of work use is writing and two‑thirds of that edits user text. ChatGPT’s user base grew to ~700M weekly users by July 2025, skewing younger and broadening gender balance. Anthropic’s Economic Index shows Claude shifting toward automation: directive, low‑interaction tasks rose from 27% to 39%, and “automation” (49.1%) now edges out “augmentation” (47%). In enterprise API traffic automation patterns appear in 77% of conversations, with a heavy coding skew (44% of API traffic vs 36% on Claude.ai). Geographic data highlight uneven adoption (Israel and U.S. hubs high per‑capita), and Anthropic warns richer regions are capturing most early productivity gains. Together the reports signal rising user trust and task delegation to LLMs—accelerating automation in developer workflows and enterprises—while raising validation, equity, and workforce‑impact questions for the AI community.