Deep Research Now Live on Gemini 2.5 Pro Experimental

Following the appointment of Josh Woodward to lead the Gemini team, Google has been making great strides in offering powerful consumer applications. Since the release of the widely praised Gemini 2.5 Pro AI model, Google has made it available to free Gemini users with some rate limit. And now, the impressive Gemini 2.5 Pro model powers the Deep Research AI agent on Gemini.

However, note that Deep Research via Gemini 2.5 Pro is only available to paid Gemini Advanced subscribers. While free Gemini users can access the Deep Research agent, it’s powered by the smaller Gemini 2.0 Flash Thinking AI model. Both are reasoning models, but 2.5 Pro has broader world knowledge, resulting in better results.

In a blog post, Google says that human raters preferred Gemini’s Deep Research by more than a “2-to-1 margin” when compared to ChatGPT’s Deep Research agent. Even in my Deep Research comparison between Gemini and ChatGPT, I leaned more towards Gemini’s output as it consistently included current information on given topics.

Bear in mind that my testing was based on the older Gemini 2.0 Flash Thinking model. Now, with the Gemini 2.5 Pro model, Google can deliver far better results. That said, there is still one major feature missing on Gemini’s Deep Research agent.

You can’t upload files when using the agent on Gemini whereas ChatGPT’s Deep Research lets you upload PDFs of paywalled science journals to expand the knowledge base. This can help the agent to incorporate valuable insights that are not freely available on the web. I hope with future updates, Google adds this important functionality to Gemini’s Deep Research agent.

Apart from that, you can now generate Audio Overviews, aka AI-generated podcasts, based on the generated report, just like NotebookLM.