Hi, I’m trying to understand how ChatGPT handles web search results during response generation.
When a response includes web search, I often see two types of links at the bottom: one labeled “Citations” and another labeled “More Results.”
I understand that Citations are the sources actually used to generate the answer—but what exactly does the “More Results” section represent?
My key questions are:
-
Are all links in the “More Results” section guaranteed to have been read (e.g., their summaries processed by the model), or could some of them have been retrieved but never actually read?
-
Is there any way for users to distinguish between:
links that were read but not used, and
links that were retrieved but never read at all?
I’ve also seen cases where the “More Results” section included 14 links.
Could it be that all 14 were actually read by the model?
If even a few were not,
→ it would suggest that “More Results” is not a list of read sources,
but rather a simple display of all retrieved search results.
ChatGPT said this question belongs in the API category, since it concerns internal system behavior and tool usage.
If that turns out to be incorrect, I apologize in advance.
Thanks in advance for any clarification!