I’ve noticed that ChatGPT often appears to automatically search the internet to provide answers, even when the web search feature isn’t explicitly activated. This behavior can be quite frustrating, especially since there’s no option to disable these automatic web searches.
I believe this undermines user control and transparency in how the AI operates. It would be helpful if OpenAI could address this by either clarifying how and when searches are triggered or by providing a straightforward toggle to disable automatic web queries entirely.
Does anyone else feel the same way, or have any insights into this issue?
“OpenAI could address this by either clarifying how and when searches are triggered”
In previous version, when web searching was back, there was a statement that ChatGPT will search the web when he struggle to provide feedback because of the lack of knowledge 'coz of knowledge cut off date October 2003.
Previously it didn’t kicks in properly sometimes, I have to force GPT to fetch current info, and after this new web search implementation he’s searchin’ too often. That’s my perspective of this.
I have the same experience. Furthermore it messes up the dialog. Instead using the dialog, it searches always for new information - not finding the correct one gfving general answers - even ChatGPT has already the correct data and knowledge.
It’s an extremely annoying behavior. Among other things, I use ChatGPT to translate article titles. It used to be great, now everytime it searches the web and gives me an answer instead of translating
And no matter how many times I say to it not to search, just translate, it seems to ignore the instructions. Now I waste double the time, since I have to regenerate the answer “Without web search”.
What can be done?