How ChatGPT serves ads (buchodi.com)

Article summary

The article explains how ChatGPT serves ads through its backend and a tracking SDK called OAIQ, which reports product views back to OpenAI. The ad platform uses Fernet-encrypted click tokens to track user interactions. The article provides technical details on how ads are injected into conversations and how the attribution loop works. The author also provides information on how to block ChatGPT ad events.

Main themes

• ChatGPT ad platform
• Ad tracking and attribution
• Local vs hosted models
• LLM inference funding
• Ad blocking and privacy

What commenters say

• The introduction of ads in ChatGPT marks the beginning of a decline in user experience and an increase in enshitification.
• Local models are a viable alternative to hosted models, offering more control and potentially better performance.
• The use of ads in LLMs raises concerns about bias and the potential for models to prioritize profit over objectivity.
• Some commenters argue that blocking ads is a simple solution, while others propose using alternative models or tools to mitigate the issue.
• The funding of free LLM inference is a challenge, with some suggesting that defense contracts or institutional subscriptions could be a solution.
• Others believe that serving ads fundamentally changes the interests of the provider, prioritizing revenue over user experience.
• There are differing opinions on the effectiveness of blocking ads, with some suggesting it is easy to block distinct ad events, while others propose using post-processing techniques to remove ads from LLM responses.
• The use of multiple models and anonymous queries may be necessary to reduce bias and ensure objective results.