Canvas (Gemini)
Last reviewed
Sources
10 citations
Review status
Source-backed
Revision
v1 · 1,500 words
Improve this article
Add missing citations, update stale details, or suggest a clearer explanation.
Suggest edit
CtrlK
Last reviewed
Sources
10 citations
Review status
Source-backed
Revision
v1 · 1,500 words
Add missing citations, update stale details, or suggest a clearer explanation.
Canvas is an interactive workspace inside the Gemini app that lets users draft and refine documents or write and run code in a panel that sits alongside the chat. Google introduced it on March 18, 2025, framing it as "a new interactive space within Gemini designed to make creating, refining and sharing your work easy."[1][2] Instead of returning a long block of text in the conversation thread, Gemini opens the output in a dedicated pane where edits appear in real time, and the user can keep talking to the model to revise the result. The feature is widely compared to ChatGPT Canvas from OpenAI and Claude Artifacts from Anthropic, all of which separate generated work from the running chat.[3]
Through 2024 and early 2025, the leading chat assistants converged on a similar idea: pull substantial generated content, an essay, a report, a piece of code, out of the linear chat and into a side panel where it can be read, edited, and iterated on without scrolling back through the conversation. OpenAI shipped ChatGPT Canvas in October 2024, and Anthropic had introduced Artifacts earlier that year. Google's answer arrived in March 2025. The aim was to turn Gemini from a question-and-answer chatbot into something closer to a lightweight productivity tool, where a draft or a prototype can be shaped over several turns rather than regenerated from scratch each time.[3]
Google announced Canvas alongside a second feature, Audio Overview, on March 18, 2025, in a post on the company's official Gemini blog.[1] The launch was covered the same day by TechCrunch, which described Canvas as a "dedicated workspace" for writing and coding projects, and noted that it was rolling out "to Gemini users worldwide."[2] Coverage from outlets including 9to5Google and ChromeUnboxed reported the same launch date and positioned Canvas as Google's interactive space for both writers and developers.[4][5]
Users reach Canvas by selecting it from the prompt bar in the Gemini app at gemini.google.com, or by going directly to gemini.google.com/canvas.[1] On launch it was available on the web; mobile support followed.
The question of who could use Canvas evolved quickly. At introduction it was tied to Gemini and Gemini Advanced access and rolled out across the languages and regions where the Gemini app is supported.[1][5] Less than two weeks later, on March 29, 2025, Google made its Gemini 2.5 Pro model (then an experimental release) available to all Gemini users, and Canvas was one of the main surfaces through which free users could exercise that model. As 9to5Google reported, Google "decided to roll out Gemini 2.5 Pro (experimental) to all Gemini users," initially on the web with the mobile apps to follow, subject to usage limits.[6] In practice this meant Canvas became reachable without a paid subscription, although the most capable model behind it carries rate limits for free accounts.[6][7] For Google Workspace customers, Canvas was made available across Business, Enterprise, Education, Frontline, Essentials, and Nonprofit tiers.[8]
Canvas opens as a panel beside the chat. When the user asks Gemini to write something or build something, the result appears in that panel rather than inline, and the conversation continues next to it. Each instruction the user sends, such as a request to shorten a section or change a button, is reflected back as an update in the panel, so the document or program evolves turn by turn while a version of the work stays visible the whole time.[1][2]
The workspace serves two broad modes, writing and coding, and later gained a menu for generating other kinds of content. Finished work can be moved out of Gemini: text can be exported to a new Google Doc, and for Workspace users, Canvas creations can also be sent to Google Slides.[1][9]
For writing, Canvas is built around iterating on a draft. After Gemini produces a first version, the user can ask for feedback, regenerate a specific section or the whole piece, and use quick editing controls to adjust tone, length, and formatting. Google's documentation and launch materials highlight highlighting a passage and asking the model to make it more concise, more professional, or more casual, then exporting the finished text to Google Docs in one click.[1][2][9]
For coding, Canvas can turn a prompt into a working prototype. Google describes generating "working prototypes for web apps, Python scripts, games, simulations and other interactive apps," and the panel can generate and preview HTML and React code directly, refreshing the live preview each time the user requests a change.[1] The interface exposes a few distinct controls: a Preview button to render an HTML or React app, a Code button to open and edit the underlying source, and a console view for inspecting errors and logs from the running preview.[9] At launch, the code preview capability was reported to be limited to the web version.[2]
The table below summarizes the two core modes.
| Mode | What it does | Typical actions |
|---|---|---|
| Writing | Draft and refine documents in a side panel | Regenerate sections, adjust tone or length, export to Google Docs[1][9] |
| Coding | Generate and run prototypes | Preview HTML or React, edit source code, view console logs[1][9] |
On May 20, 2025, Google expanded Canvas under the banner "More ways to create," adding a Create menu that turns text into several kinds of dynamic output.[8] The new options were:
| Output type | Description |
|---|---|
| Web pages | Pages with interactive elements such as information cards and charts[8] |
| Infographics | Visual layouts that condense complex information into a readable format[8] |
| Quizzes | Practice assessments generated from an uploaded document or a prompt[8][10] |
| Audio Overviews | Podcast-style discussions between two AI hosts[1][8] |
Audio Overview, launched on the same day as Canvas in March 2025, uses the same underlying technology as the Audio Overviews in NotebookLM: it generates a spoken, podcast-style conversation between two AI hosts that summarizes uploaded material and draws connections across it.[1] When folded into the Canvas Create menu in May, Google said Audio Overviews could be produced in more than 45 languages.[8]
The quiz feature, detailed in a separate Workspace announcement on May 20, 2025, lets a user generate an interactive quiz on a subject by uploading source documents or simply typing a request such as a practice quiz on a given topic. The quiz returns multiple-choice or true-or-false questions with instant feedback and hints, and reports an overall score, an accuracy breakdown, and notes on areas to improve.[10] Several of the May additions are tied to Gemini's 2.5-generation models, and Google noted age and sharing restrictions for Workspace accounts, including that the relevant rollouts targeted users 18 and older and that business and education users could not share Canvas-created content even though they could view links shared from personal accounts.[8][10]
Canvas is one of three closely related features from the major assistant makers. TechCrunch noted at launch that Canvas is "similar in concept" to ChatGPT Canvas and to Claude Artifacts, each of which gives an AI chatbot a separate workspace so that documents and code can be edited with more precision than a plain text reply allows.[2][3]
The implementations differ in emphasis. ChatGPT Canvas leans toward writing and inline text editing, with suggestion buttons for quick changes. Claude Artifacts tends to render interactive output, including code and HTML or SVG visuals, directly inside the conversation and can publish artifacts to the web. Gemini Canvas is distinguished chiefly by its ties to Google's own products: text and slides can be exported to Google Docs and Google Slides, and the workspace draws on Google models such as Gemini 2.5 Pro for code generation and reasoning.[3][6][9] All three reflect the same shift, from chatbots that only answer to tools that help produce a finished artifact over multiple turns.