Gemini adds an important piece to its transformation from a conversational assistant into an operational workspace for content creation. Google's app can now generate, preview, and edit documents, PDFs, spreadsheets, presentations, Markdown files, TXT files, and code directly within the conversation. Users no longer need to export a file, open it in another application, and head back to the chat to ask for corrections: the document remains at the center of the interaction with the model.
The feature is available to all Gemini users in the countries and languages where the app is supported. However, access is not identical across all devices. PDF viewing and editing are rolling out to both the web and the mobile app; for Google Docs, Google Sheets, slide decks, Markdown, TXT, and code files, users currently need to use the web version of Gemini.
The update may seem like a natural extension of file generation, a capability Google had already introduced in the app in recent months. The substantial difference lies in the last mile: the output is no longer just a downloadable attachment, but an editable object right inside the interface where you are working with AI.
From prompt to polished file, without switching between apps
The new editor expands to full screen when manual intervention on the content is needed. Users can write, edit, and fine-tune the file directly, alternating between their own edits and Gemini's suggestions. It is also possible to select a passage and prompt the model to work on that specific section, rather than giving it broad instructions covering the entire document.
This approach aims to provide greater control over a common chatbot limitation: a prompt like “make the text clearer” risks generating sweeping, often unwanted rewrites. By selecting a sentence, paragraph, or visual element, the scope of the edit becomes much more contained. Among potential examples, Google highlights converting a selected table into a chart.
In the case of PDFs, Gemini can also be used as a processing step after uploading existing material. One use case shown by Google involves handwritten lecture notes: users can generate a study guide in PDF format, manually review its text, highlight sections to request changes on, and follow up with further questions about the content. It is therefore not just about creating a file from scratch, but about combining user-provided material, generation, and revision.
What changes compared to Workspace
Google already offers Gemini features inside Workspace applications, including Docs and Sheets. However, that model starts from the productivity app: the user opens a document and summons the assistant within the context of the document itself. The update announced for the Gemini app reverses this workflow: the entry point is the conversation, from which the file is born or refined.
It is a practical distinction for those who use AI primarily to turn a prompt into an output ready to be discussed or revised: a draft presentation, a spreadsheet, a study outline, a structured document, or a code file. The chat retains the context of previous instructions, while the editor allows users to visually inspect the result before downloading it or moving to another tool.
This move brings Gemini closer to a category of services where the chatbot is no longer just a text box, but a production surface. Anthropic, for instance, has already introduced tools in Claude to create and edit documents and presentations during conversations. Competition among generative assistants is shifting to this battleground as well: producing plausible text is no longer enough; it must be delivered in a format that seamlessly fits into daily workflows.
Supported formats and the boundaries of mobile
The list covers the most common file types in individual productivity and professional material preparation: PDFs, Google Docs, Google Sheets, presentations, Markdown files, plain text, and code. However, the varying support across platforms signals that the rollout does not yet turn the phone into a complete replacement for the browser for this type of task.
On mobile, the experience focuses on PDFs, arguably the most useful format for reviewing, annotating, or quickly processing received materials. Working on documents, sheets, slides, and code, however, still requires switching to the web. It is a significant limitation for tasks requiring composition, layout, and more extensive controls: creating a presentation or adjusting a complex table remains, at least for now, a task designed for larger screens.
Furthermore, the integrated editor does not eliminate the need to carefully check the output produced by the model. A well-formatted file can still contain inaccurate summaries, flawed formulas, misinterpreted data, or stylistic changes unsuited to the intended audience. This is especially true for PDFs derived from notes, spreadsheets, and materials intended for educational or professional settings. The ability to edit the document directly makes verification more convenient, but not superfluous.
An interface that tries to keep work within the chat
For Google, the strategic advantage is clear: reducing friction between Gemini and external tools makes the app more useful even after the initial response. Until now, a conversation might have served to generate a draft, an analysis, or an outline, while the refining work often happened elsewhere. With previews, contextual selections, and full-screen mode, Gemini aims to keep that phase within the same experience.
For users, the benefit will depend on the quality of control offered over individual elements and how faithfully Gemini maintains structure, formatting, and editorial intent during revisions. The value lies not in automating every single edit, but in being able to choose when to delegate and when to step in manually without breaking the flow.
General availability makes the update immediately accessible, but its real-world adoption will be measured in less spectacular use cases: reviewing a report, cleaning up a spreadsheet, turning notes into readable material, tweaking a slide, or making a targeted edit to a PDF. These are repetitive tasks, and precisely for that reason, they can define whether a generative assistant truly becomes part of everyday productivity tools.
