Google’s AI-powered research tool, Gemini Deep Research, is getting a significant upgrade allowing users to upload their own files for analysis. This upcoming capability, first spotted by TestingCatalog, fundamentally changes how the feature operates, enabling more personalized and tailored research insights grounded directly in a user’s private documents and data.
So far, Deep Research primarily relies on scanning the open web. The enhancement brings Gemini’s research functionality closer to Google’s dedicated AI note-taking tool, NotebookLM, which is built specifically around processing user-provided source materials.
The file upload feature will be accessible on both the web and mobile versions of Gemini Apps. Users can tap the plus sign when prompting Gemini and select “files” to upload documents. Supported file types are extensive, including plain text, various code files, document formats like DOC, PDF, and Google Docs, as well as tabular data and spreadsheets. This broad compatibility allows users to analyze a wide range of personal or work-related content.
Expanding Research Capabilities
With file upload, Gemini Deep Research moves beyond general web summaries to synthesize information directly from user-provided sources. This mirrors a core strength of NotebookLM, which is designed to work exclusively with uploaded materials, generating summaries, answering questions grounded in those sources.
NotebookLM has steadily added features, including Audio Overviews in over 50 languages and Mind Maps for visual organization. While NotebookLM makes copies of local files and doesn’t automatically sync changes, it supports various inputs including image-only PDFs and YouTube transcripts.
Google recently rolled of file uploads to the free Gemini tier for normal model requests following its initial introduction for paid Gemini Advanced subscribers in May 2024. Gemini Advanced users benefit from a larger context window, capable of handling up to 1,500 pages or 30,000 lines of code for file uploads, as noted in the Gemini Apps release notes. The feature is also supported within the instructions for creating personalized Gemini Gems.
Privacy Considerations
As users upload personal or potentially sensitive documents, data handling practices become crucial. According to the Google Help documentation, even when Gemini Apps Activity is turned off, conversations and uploaded files will be saved with your account for up to 72 hours.
If activity is enabled, data is stored for 18 months by default, though this can be adjusted. Google states that human reviewers may read and annotate conversations, disconnected from the user’s Google Account, and this reviewed data can be kept for up to three years and used for model training. Google advises users not to “enter confidential or sensitive information when using Gemini Apps.”
Competitive Landscape
The addition of file upload capabilities intensifies competition in the AI research tool market. Google previously made its Deep Research feature free for all users in March 2025, challenging tools like OpenAI’s ChatGPT Deep Research and Perplexity AI. OpenAI has also expanded ChatGPT’s capabilities, adding a lightweight Deep Research option recently and connectors to external data sources like GitHub repositories.However, integrating broad file upload support directly into a widely accessible tool like Gemini provides a distinct advantage for users needing to analyze their own diverse datasets.
