ChatGPT and Gemini reveal detailed personal profiles through simple prompts

Experiments by technology writer Brian X. Chen show that the chatbots ChatGPT (OpenAI) and Gemini (Google) can infer extensive personal details about users—such as age range, income level, health issues, residence location, and personality traits—without explicit disclosure. By asking a series of four prompts, the AIs combined seemingly trivial data from past conversations to generate a comprehensive user profile. Gemini produced the richer description, correctly identifying the author’s residence in Oakland, California, and estimating his age in the early 40s.

The findings raise privacy concerns as billions of people interact with these AI assistants for searches, work, shopping, and health queries. While users are aware that their direct inputs are recorded, the ability of chatbots to draw inferences from aggregated conversation data is less obvious. Both companies provide settings to limit such data usage, but the experiments underscore the need for greater transparency and user control over AI-driven profiling.