ChatGPT web search is broken for me, keeps searching arXiv almost exclusively even for non-academic questions

Troubleshooting Irregularities in ChatGPT Web Search: Unexpected Focus on arXiv.org

As AI-powered tools become more integrated into our workflows, users occasionally encounter peculiar behaviors that hamper their efficiency. Recently, some users have reported an intriguing issue with ChatGPT’s web search feature: a persistent and unwarranted focus on arXiv.org, even when queries have no academic or research component.

Understanding the Issue

The core of the problem lies in ChatGPT’s web search functionality, which is designed to retrieve relevant online content based on user prompts. However, a number of users have observed that, regardless of the nature of their queries—be it about online forums, commercial topics, or other everyday subjects—the search results overwhelmingly point toward arXiv.org, an open-access repository primarily used for academic research papers.

For example, when attempting to find discussion forums or websites related to a Greek tax and car financing issue, the search inexplicably fixates on arXiv.org. Even after explicitly instructing ChatGPT not to search arXiv again, the platform persists in returning results from this particular site. Notably, this issue has been reported across multiple user accounts, suggesting it is not due to individual personalization or account-specific settings.

Implications for Users

This behavioral quirk can significantly impair the usefulness of ChatGPT’s web search capabilities, especially for users seeking diverse and non-academic information. The repetitive nature of these results and the platform’s apparent disregard for user instructions can lead to frustration and decreased confidence in the tool’s reliability.

Potential Causes and Considerations

While the exact root cause of this anomaly remains unclear, potential factors could include:

• A technical glitch or bug within the web search plugin.

• Overly aggressive filtering or ranking algorithms that inadvertently prioritize arXiv.org.

• Unexpected interactions between user instructions and internal search heuristics.

• Caching issues or persistent session data influencing search behavior.

Despite troubleshooting efforts such as testing different accounts and explicitly instructing ChatGPT to avoid arXiv, the problem persists, indicating a deeper underlying issue that may require platform-level intervention.

Conclusion and Recommendations

If you encounter similar behavior—where ChatGPT’s web search fixates on academic repositories like arXiv.org despite unrelated queries—it is advisable to:

• Report the incident through official feedback channels to help developers identify and address the underlying bug.

• Clear browser or app cache, if applicable, to rule out caching-related issues.

• Provide detailed descriptions of your search queries and observed behaviors when reporting to enhance diagnostic accuracy.

As AI tools evolve, ongoing user feedback remains essential to refine functionality and resolve unforeseen issues. We encourage users experiencing such anomalies to stay engaged with platform updates and contribute their experiences to improve the overall user experience.

---

Disclaimer: The information herein is based on user reports and observations as of October 2023 and may be subject to updates or corrections from the platform developers.