Researchers Say ChatGPT Builds Detailed Psychological Profiles of Users

OpenAI's research finding that ChatGPT retains and interprets user inputs to assemble detailed psychological profiles, as presented by a coalition of German, Indian and Dutch scientists at the ACM Web Conference 2026, raises immediate privacy and compliance concerns for website owners who embed the model. For site operators, this means any visitor interaction with an embedded ChatGPT widget could be silently harvested into a nuanced user persona, potentially exposing personal data under GDPR, India’s Personal Data Protection Bill, and the EU AI Act. If your site collects consent for analytics but not for deep behavioral profiling, you could face regulatory scrutiny, especially when the model is used for marketing, support, or content personalization. Moreover, search engines may downgrade pages that host AI that performs undisclosed profiling, impacting SEO and traffic. **Actionable steps:** 1. **Update your llms.txt** – explicitly declare that the embedded LLM performs user profiling and list the data categories processed; include a “no‑profile” flag for bots that should be excluded. 2. **Implement granular bot detection** – use llmscentral’s traffic analytics to differentiate between human visitors and AI‑driven crawlers, then block or sandbox profiling requests from non‑human agents. 3. **Add a double‑opt‑in consent layer** before any ChatGPT interaction, storing the consent flag in a cookie that your llms.txt references, ensuring you can prove compliance if audited.