OpenAI Introduces ChatGPT Deep Research

OpenAI has launched a new AI feature. This tool helps users do detailed and complex research on the ChatGPT platform. This new feature, called “Deep Research,” helps professionals in finance, science, policy, and engineering. They need accurate and detailed information.

Additionally, it could prove beneficial for consumers making well-informed purchasing decisions on items like cars, appliances, and furniture.

Unlike standard chatbot interactions that provide quick responses, Deep Research is tailored for users who need to analyse multiple sources and meticulously assess information before drawing conclusions.

To utilise Deep Research, users can select the option within ChatGPT’s composer and enter their query, with the ability to attach relevant files or spreadsheets. Initially, the feature is web-only, with mobile and desktop app integration slated for release later this month. Response times range from five to 30 minutes, and users will receive notifications upon completion.

Features and Future Enhancements

Currently, Deep Research provides text-based outputs. However, OpenAI intends to enhance the experience by incorporating embedded images, data visualisations, and other analytical tools. Future updates will also enable the integration of specialised data sources, including subscription-based and internal resources.

Accuracy and Reliability Concerns

Given that AI models are prone to errors, OpenAI emphasises that all Deep Research responses will include full documentation, citations, and summaries to help users verify information. Nevertheless, concerns remain about AI-generated misinformation, as ChatGPT’s existing web search tool has previously provided inaccurate results.

To enhance precision, OpenAI has integrated a specialised version of its newly developed o3 “reasoning” AI model, which has been trained using reinforcement learning.

This model is optimised for web browsing and data analysis, allowing it to interpret vast amounts of text, images, and PDFs while adapting based on the information it encounters.

Additionally, it can analyse user-uploaded files, generate and refine graphs using Python, and cite specific passages from its sources.

Despite its strengths, OpenAI acknowledges that Deep Research has limitations. It may sometimes struggle to distinguish authoritative sources from misinformation and fail to convey uncertainty when appropriate. Additionally, it can occasionally introduce formatting errors in reports and citations.

The Future of AI-Assisted Research

For students, researchers, and professionals, a well-cited, in-depth AI research tool is likely to be more appealing than a basic chatbot summary lacking references.

However, the true test will be whether users rigorously analyse and fact-check the provided information or simply rely on it at face value.

Interestingly, OpenAI is not the first to introduce such a feature. Google announced a similar AI-powered Deep Research tool just two months ago, highlighting the growing competition in AI-driven knowledge work.