Google’s Gemini 2.5 Pro now powers deep research: What might the impact be for online R&D research?

Pundits had projected 2025 to be the year of the AI agent. And while the results to date have been mixed, Deep Research from OpenAI has been one of the most visible and popular AI agents. Give it an online research task. It asks some clarifying questions, and it then proceeds to do a deep dive on a subject of your choice. Wait about 10 to 30 minutes, and you get the results – sometimes pretty good, but occasionally laced with a couple AI ‘hallucinations.’ All in all though, Deep Research can do a first wave pass at research that might take several times longer to do the old-fashioned way.

Now, Google, whose Gemini 2.5 Pro model is among the strongest around based on benchmarks and informal user reviews, is giving OpenAI a run for its Deep Research money. Its own Deep Research feature, now supercharged by its latest flagship model Gemini 2.5 Pro, signals a potential shift, offering the public and R&D professionals alike a new tool that could streamline everything from basic online information synthesis to complementing more specialized AI tools from academic publishers.

So what’s the upshot for the folks in the lab coats or managing research budgets? Well, Google’s play seems to be about bringing this kind of AI research assistant power to more people, potentially at a lower price point than some alternatives, tucked into their Gemini Advanced subscription. For R&D teams, this could mean getting a decent first draft of a competitive landscape analysis, a summary of recent patent filings in a specific area, or even a basic literature review pulled together much faster. It can scour dozens or even hundreds of websites as it researches.

Google’s internal tests claim users preferred their results 2-to-1 over OpenAI’s, but take that with a grain of salt – company benchmarks always look good. But while OpenAI reportedly limits access to its inference-hungry deep research tool to a small number of queries per month for many paid users (those paying $200 per month get more access), Google is apparently offering more liberal access at a $20 per month price point. The cap is reportedly around 20 deep dives per day.

In terms of output, there is some degree of AI fluff to wade through. (Full sample report here.) A test query on “advances and challenges in solid-state battery development” had the following sentence — with a citation nonetheless: “The relentless pursuit of safer, more energy-dense, and longer-lasting energy storage solutions has propelled solid-state batteries (SSBs) to the forefront of electrochemical research and development. At the heart of this transformative technology lies the solid-state electrolyte (SSE), a material class poised to redefine battery architecture and performance.” Besides that, the research is generally well organized and grounded with references.

Also, the process from Google’s Deep Research is arguably more user friendly. While the OpenAI version of the tool offers a handful of clarifying questions to ensure that its research aligns with what you want, Google’s approach is more hands-off — it comes up with a plan and allows you to tweak it or proceed with its default approach. There’s even a ‘Showing Thoughts’ feature if you want a peek behind the curtain, revealing steps like identifying the scope or brainstorming concepts as it works.

For those tired of reading dense reports, it can whip up an ‘Audio Overview,’ basically turning the text into a two-voiced podcast summary based on technology from Google’s NotebookLM.