Last week, OpenAI implemented notable enhancements to the web search capabilities of ChatGPT. The San Francisco-based company has made strides in improving both the quality of responses and the chatbot’s ability to adhere to user instructions. Users can now submit longer and more intricate queries, while the chatbot has gained the ability to perform image-based searches, retrieving information from publicly available websites based on object images. Additionally, the features in ChatGPT’s Projects function have been expanded, and the Canvas tool is now available for download.
Web Search in ChatGPT Receives Significant Upgrades
In a comprehensive changelog released on June 13, OpenAI outlined numerous enhancements to its web search feature, now accessible to users. The integrated search tool was made available to all users in October 2024, and since then, OpenAI has actively worked on expanding its functionality and increasing its user capacity.
The latest updates enable ChatGPT to deliver more current and thorough information. A key focus has been on enhancing response quality, as the web search tool now demonstrates improved comprehension of user intent, allowing responses to be more accurately tailored. Furthermore, it can manage longer conversational contexts, meaning that even if an initial user prompt isn’t explicit, ChatGPT can extract relevant information from previous dialogue leading up to the query.
This capacity is partly due to an upgraded memory function that OpenAI has been incorporating into various ChatGPT features. The Projects component now references past projects in conversations, enhancing context and user experience.
Additionally, the web search function has improved its instruction-following capabilities. This enhancement is particularly noticeable in longer interactions, where the chatbot can refine the scope of a query based on ongoing user input, minimizing repetitive answers.
Another significant update allows the chatbot to conduct multiple searches for a single query, enabling it to gather more information by exploring related topics when needed.
Moreover, ChatGPT now has the ability to conduct web searches using images uploaded by users. Testing from Gadgets 360’s staff confirmed that the chatbot successfully identified objects in images and provided relevant responses to accompanying queries.
OpenAI has also acknowledged existing limitations with the current functionality. Users may experience delays while waiting for responses, and at times, the AI may employ a transparent chain-of-thought (CoT) approach for relatively simple queries. CoT typically suggests that the model is applying reasoning, which is usually reserved for more complex inquiries. The company is expected to address these concerns in future updates.
