Konitzny: ChatGPT user data in shopping feature? Traces of potential personalization

Key Takeaways:

• Analyzing the new ChatGPT Shopping Feature Konitzny found data from his previous interactions that could be used for personalizing results:

  1. Model Usage Breakdown

  2. Message Length & Conversation Depth

  3. Topics, Language & Model Breakdown

  4. Device Settings & Time Since Load

  5. User Agent & Location

  6. Account Settings: Time zone, plan, and activities

• How to test yourself: Look for the ID in the URL where you used the shopping assistant. Then, go to DevTools > Network and search for that ID. The rest is in the code. This data is only tracked if ChatGPT memory is turned on.

• Segonzac points out that this sort of data was used in the ChatGPT usage statistics published by OpenAI

Sources:

• https://www.linkedin.com/posts/davidkonitzny_how-much-chatgpt-actually-knows-about-me-ugcPost-7398991088230854656-kQ95/

• https://www.linkedin.com/posts/davidkonitzny_user-data-summarized-using-the-new-openai-activity-7399032003947343872-hk-0/

• https://cdn.openai.com/pdf/a253471f-8260-40c6-a2cc-aa93fe9f142e/economic-research-chatgpt-usage-paper.pdf