Google is pushing Gemini further into research work with a new feature called Deep Research. The tool is built to take a user’s question, examine relevant information across the web and return a structured brief that can be edited later in Google Docs.
The launch gives Gemini a more active role than a standard chatbot response. Instead of answering only from a prompt, Deep Research is designed to act more like a research assistant that gathers material, summarizes it and links back to original sources.
What Deep Research does inside Gemini
Deep Research is part of Gemini Advanced, the more capable version of Google’s chatbot platform. Access is limited to users on the Google One AI Premium Plan, which is priced at $20 a month.
Google says the feature uses “advanced reasoning” and “long context capabilities” to build research briefs. In practice, that means Gemini can work through a broader research question, look for relevant web information and organize the results into a report-style format inside the Gemini apps.
The finished briefs are not meant to remain locked inside the chat window. Users can export them to Google Docs, where they can continue editing, revising or expanding the work.
Each brief includes summaries and links to the original source material. That matters because it gives users a path back to the web pages Gemini used, rather than presenting the report as a standalone answer with no visible trail.
Where users can try it first
Deep Research is not available everywhere in Gemini at launch. It is available only in English on desktop and the mobile web to start.
Users who have access can find it by selecting the “Gemini 1.5 Pro with Deep Research” option from the model drop-down menu. Google says the feature will come to the Gemini mobile apps in early 2025.
David Citron, product director for the Gemini apps, described the system as a way to guide Gemini’s browsing and research using Google’s ability to find relevant web information. In the blog post provided to TechCrunch, he wrote: “Deep Research saves you hours of time.”
That is the central promise of the feature: reduce the early work of gathering sources, sorting them and turning them into a usable brief. For many users, that first pass is the slowest part of research, especially when a topic requires checking multiple sources before writing can begin.
The education concern is not just accuracy
Deep Research arrives with a familiar warning attached: AI systems can make mistakes and hallucinate. A polished report from a chatbot can still contain errors, and users need to check the underlying sources before treating the output as reliable.
But the concern around education goes beyond incorrect information. The source article points to the risk that students may use generative AI to outsource brainstorming and writing, not only to speed up research.
Jessica Grose, writing in an op-ed in The New York Times, described students increasingly leaning on generative AI for work that would otherwise require them to think through ideas and deal with difficult writing tasks. The concern is that convenience can weaken the habits that research assignments are supposed to build.
The source also notes at least one study linking heavy usage of ChatGPT among students to higher levels of procrastination, memory loss, and lower grade point averages. Deep Research could make that tradeoff sharper because it handles more of the research workflow in one place.
Used carefully, a research brief can help someone find starting points. Used carelessly, it can become a substitute for the harder work of reading, comparing sources and forming an argument.
Publishers face a tougher web traffic question
Deep Research also raises a business concern for publishers. If Gemini gathers information from websites and turns it into a brief, users may feel less need to visit the pages that provided the material.
That could affect ad revenue for sites that depend on readers arriving from search and other discovery channels. The source article compares the risk to Google’s AI Overviews, the AI-generated summaries that appear for certain Google Search queries.
According to one source cited in the article, publishers have seen a 5% to 10% decrease in traffic from search since AI Overviews launched early this year. On the revenue side, an expert cited by The New York Post estimated that AI-generated overviews could lead to more than $2 billion in losses for publishers.
Google often says it works closely with publisher partners, respects paywalls and allows websites to block its AI scraping at the domain level. Even so, outlets can face a difficult choice: allow scraping or risk losing visibility in Google Search, at least for now.
Google argues that Deep Research could “connect users to relevant websites they might not have found otherwise so they can dive deeper to learn more.” Whether that happens depends on how users behave after reading a Gemini brief. If they click through and read the source material, the feature could support discovery. If they stop at the summary, it could pull attention away from the open web.
Gemini 2.0 Flash is arriving too
Deep Research is not the only Gemini update in the source article. Google is also giving both free and paying Gemini users access to Gemini 2.0 Flash, its newest flagship AI model.
The version being introduced is experimental and optimized for chat. The full version will arrive in January.
Users can select 2.0 Flash from the Gemini model drop-down on desktop and the mobile web, though not yet in the mobile apps. Google says it should bring better performance across a number of tasks and faster responses.
There is one limitation: Google cautions that some Gemini features “won’t be compatible with [the] model in its experimental state.” The company did not specify which features are affected.
Together, Deep Research and Gemini 2.0 Flash show Google moving Gemini in two directions at once: more capable research workflows for paying users, and faster chat performance for a wider group of users. The bigger question is how much of the web’s research process should be compressed into an AI-generated brief.