Google has recently released a new AI tool named Deep Research, which utilizes its Gemini robot to assist users in conducting in-depth research on the Internet. Currently, this tool is only available to subscribers of Gemini Advanced and only supports English-speaking users. Users can give simple instructions to Gemini to conduct research on specific topics and generate a detailed research report.

When using Deep Research, users only need to tell Gemini the topic they want to research, and the system will create a "multi-step research plan" that users can choose to edit or directly approve. Gemini's working method is to first look for relevant information on the Internet and then conduct relevant searches. This whole process will be repeated multiple times to ensure the comprehensiveness and accuracy of the research.

After completing the research, Gemini will provide a report containing "key findings" along with links to the websites where it found the information. Users can also ask Gemini to expand or adjust certain parts and export the generated research report to Google Docs. This process is somewhat similar to the page function provided by the AI search engine Perplexity, which can generate customized web pages according to user prompts.

The release of Deep Research is part of Google's overall rollout of the Gemini 2.0 model and marks Google's progress in the era of "agentic" AI. Agentic AI systems can perform specific tasks on behalf of users, and Deep Research is an embodiment of this trend. Meanwhile, Google also announced that it will launch Gemini Flash 2.0, a faster next-generation chatbot that will be open to developers.

Currently, Deep Research is only available to subscribers of Gemini Advanced. Users can experience it by accessing Gemini and selecting "Gemini 1.5 Pro with Deep Research" in the model dropdown menu.

Official Blog: https://blog.google/products/gemini/google-gemini-deep-research/