OpenAI is changing how some users work with ChatGPT by adding a project space beside the standard chat window. The new interface, called canvas, is designed for writing and coding tasks that need more than a single back-and-forth exchange.
Instead of treating every request as a fresh chat response, canvas gives users a separate window where generated text or code can be reviewed, highlighted, revised and extended. The shift matters because many useful AI outputs are not finished products on the first try. They are drafts, starting points or blocks of code that need targeted changes.
A Workspace Beside The Chat
Canvas opens next to the normal ChatGPT conversation and acts as a working area for longer outputs or more complex projects. Users can generate writing or code directly inside that space, then select specific sections and ask the model to edit only those parts.
That is the core difference from a standard chatbot exchange. In a regular chat, fixing a weak section often means rewriting a prompt, explaining what went wrong and asking for a new version. With canvas, the user can point to the exact material that needs attention.
OpenAI is rolling out canvas in beta to ChatGPT Plus and Teams users on Thursday. Enterprise and Edu users are set to receive access next week. OpenAI also says it plans to make the feature available to free users once canvas is out of beta.
Why Editable AI Workspaces Are Gaining Ground
The move reflects a broader pattern among consumer AI providers. Several companies are building editable workspaces as a more practical interface for generative AI, especially when the task involves writing, programming or other multi-step work.
The source article compares ChatGPT canvas with Anthropic's Artifacts, which launched in June, and the coding companion Cursor. OpenAI is working to match rival offerings while also adding new capabilities to ChatGPT as it tries to grow its paid user base.
The reason is straightforward: AI chatbots today cannot complete large projects from a single prompt. They can, however, often produce a useful first version. A workspace like canvas is meant to turn that first version into something easier to refine.
For users, the practical benefit is control. Rather than accepting a full answer or regenerating everything, they can work in smaller pieces. That makes the AI interaction feel less like submitting a request and more like editing a document or reviewing code with an assistant.
How Canvas Works For Writing
In a demo with TechCrunch, OpenAI product manager Daniel Levine showed how canvas could help with an email. A user can ask ChatGPT to draft the email, and the generated text appears in the canvas window.
From there, the user can adjust the length with a slider, making the writing shorter or longer. They can also highlight individual sentences and ask ChatGPT to change them. The source gives examples such as asking the model to make a sentence sound friendlier or to add emojis.
Canvas can also rewrite the whole email in another language while preserving the user's working context. That makes the feature useful for common writing tasks where tone, length and wording often matter as much as the basic message.
The writing tools show the larger direction of the interface. ChatGPT is not only producing text; it is giving users controls for shaping the output after it appears. That is especially relevant for emails and other short documents, where a small wording change can affect how the message lands.
How Canvas Works For Coding
The coding version of canvas includes different controls. In the TechCrunch demo, Levine asked ChatGPT to create an API web server in Python, and the result appeared inside the canvas window.
One button, called “add comments,” lets ChatGPT insert in-line documentation that explains the code in plain English. Users can also highlight a section of code and ask ChatGPT to explain it or answer questions about it.
ChatGPT is also getting a “review code” button. That feature can suggest specific edits for code in the window, whether the code was generated by ChatGPT or written by the user. The user can approve the suggestions, edit them manually or decline them.
If the user approves a suggested fix, ChatGPT will try to repair the bugs itself. That workflow keeps the user in the decision-making role while giving the model a clearer place to operate than a long, repeated chat thread.
What This Means For ChatGPT Users
Canvas suggests that the future of chatbot interfaces may be less about one-shot answers and more about shared workspaces. For writing, that means faster revision of drafts. For coding, it means a clearer way to inspect, explain and improve generated code.
There are also multiple ways to open it. In the demo, Levine selected “GPT-4o with canvas” from the ChatGPT model picker. OpenAI says canvas windows will also appear automatically when ChatGPT detects that a separate workspace could help, such as with longer outputs or complex coding tasks. Users can also type “use canvas” to open a project window.
The feature is still in beta, and its early availability is limited to paid ChatGPT users before expanding to Enterprise and Edu users next week. But OpenAI's plan to offer canvas to free users after beta indicates that the company sees the interface as more than a niche tool.
For anyone using AI to draft, revise, review or code, the key change is the same: canvas gives ChatGPT a place to work on the artifact itself. That does not make the model a complete project finisher, but it may make the path from rough output to usable result more direct.