Dear OpenAI Team,
I’ve encountered a recurring issue with ChatGPT’s search functionality. Despite its capability to perform web searches, it often fails to activate the tool even when explicitly requested. This leads to inefficiencies and unnecessary back-and-forth to get the tool to respond correctly.
Steps to Reproduce:
-
Start a new ChatGPT session.
-
Request a search for up-to-date information (e.g., “Search the internet for recent updates on [topic]”).
-
ChatGPT responds by stating it doesn’t have internet access, even though the feature is available in the current session.
-
After repeated prompts, ChatGPT eventually performs the search.
Impact:
This behavior disrupts workflow, creates frustration, and reduces the utility of ChatGPT as a tool for timely and accurate information.
Suggestion for Improvement:
Ensure the search functionality consistently activates when needed without requiring repeated input.
Improve transparency about the tool’s current availability or limitations.
Thank you for your attention to this matter.
Sincerely,
Zach Banks
I am a pro user
Chateau Fails to Use Web Search Despite Clear User Instruction
Issue Summary:
When I requested ChatGPT to perform a web search, it initially claimed it could not perform the requested action, even though the search functionality was enabled. Only after repeated prompts and evidence provided by me (including a screenshot) did it finally proceed to execute the search as originally requested.
Steps to Reproduce:
- I directly and explicitly instructed ChatGPT to perform a web search for information.
- ChatGPT incorrectly stated it could not perform the action and suggested alternative solutions (e.g., government websites), despite the web search capability being functional.
- I reiterated the request and uploaded a screenshot proving the discrepancy between its stated functionality and its actual capabilities.
- I directly commanded ChatGPT to “do it now,” at which point it hesitated again, asking for clarification.
- After continued insistence, ChatGPT finally initiated the web search and successfully returned results, confirming that the initial refusal was incorrect.
Expected Behavior:
ChatGPT should have immediately performed the web search upon my first request, without unnecessary delays or incorrect claims about its capabilities.
Actual Behavior:
It took multiple prompts, escalating commands, and evidence of functionality for ChatGPT to complete the requested action.
Impact:
This behavior wastes time, creates frustration, and undermines trust in the system’s ability to fulfill straightforward requests. It also demonstrates inconsistency in tool execution.
Recommendation:
Ensure that ChatGPT consistently prioritizes explicit user commands to use enabled tools like web search, avoiding defaulting to inaccurate responses.