Last week, OpenAI made significant enhancements to the web search capabilities of ChatGPT. The San Francisco-based AI company has improved response quality and instruction-following abilities, allowing the tool to manage longer and more complex queries effectively. Additionally, there is now an intriguing feature that enables the chatbot to perform image-based web searches, where it can analyze an object from a photo and retrieve information from publicly available websites. Furthermore, OpenAI expanded the functionality of ChatGPT’s Projects feature and made the Canvas feature downloadable.
Web Search in ChatGPT Receives Upgrade
In a changelog published on June 13, OpenAI outlined a series of enhancements to the web search functionalities now accessible to users. The native search engine was launched for all users in October 2024, and the company has been steadily increasing its capabilities and rate limit since that time.
The latest update means ChatGPT’s web search can provide users with more timely and extensive information. Emphasis has been placed on improving response quality, enabling the tool to better grasp the intent behind user queries and tailor its answers accordingly. Additionally, the tool is capable of remembering longer conversational contexts, allowing it to extract relevant information even if the user’s initial prompt is incomplete.
This improvement stems from the recently implemented memory function that OpenAI has began integrating into various ChatGPT features. A notable example is the Projects feature, which can now reference past projects during discussions.
Moreover, the web search feature is now more proficient in following user instructions, particularly in extended conversations. This improved capability allows the chatbot to refine the focus of queries based on the user input, helping to minimize repetitive answers.
Another appealing upgrade is the ability of the tool to conduct multiple searches for a single query. This new feature enables the chatbot to retrieve additional information by running searches on related topics when required.
In addition, ChatGPT can now perform web searches based on uploaded images from users. Staff members at Gadgets 360 tested this feature and confirmed that the chatbot accurately identified objects in images and addressed queries associated with them.
While these advancements are notable, OpenAI has acknowledged existing limitations within the feature. Users may experience longer wait times for responses, and in some cases, the AI might default to using transparent chain-of-thought (CoT) reasoning for simple queries. CoT refers to the model’s reasoning capability, which is usually reserved for more complex inquiries. OpenAI noted plans to address these issues in the future.
