Web Search Update Has Compromised ChatGPTs Conversational Depth and Knowledge Retention

Since the introduction of web search, my experience with ChatGPT has significantly declined. Previously, it felt like a true conversation where I didn’t have to repeat myself, as it retained context across responses. Now, each answer feels isolated, treating every question as if it’s a standalone query, even within the same conversation window, with search deactivated.

Additionally, it frequently defaults to web search for simple, daily questions, even when I’d prefer it didn’t. This often results in repetitive, unhelpful answers.

For instance, when I asked for dividend data on an index over the past 10 years, it couldn’t find the information. I tried rephrasing the question, switching models from GPT-4o to GPT-4.0, and eventually instructed it to avoid web search and rely solely on its internal knowledge. Suddenly, it provided the correct information in a table. However, when I repeated the request, it once again claimed it didn’t know.

ChatGPT is inherently a powerful tool, but since the web search update, it feels like a bug has disrupted its ability to think independently. It often defaults to citing sources, even when it already knows the answers, leading to responses that lack the depth and continuity I once appreciated.