What Is ChatGPT Canvas, and Is It Still Available?

If you have gone looking for ChatGPT Canvas recently and could not find it, you are not missing a setting. OpenAI retired it.

That is the part most guides still get wrong. Search for Canvas today and you will find a wall of tutorials, many of them dated this year, explaining how to open a feature that is no longer there on current models.

Canvas was a genuinely good idea, and the thinking behind it did not disappear — it got absorbed into how ChatGPT replies by default. So the useful version of this question is not just "what was it", but "what do I use instead".

This guide covers what Canvas actually did, when and why it went away, what took over its job, and how to move any work you still have sitting in it.

The Short Answer

Canvas was a side-by-side editing workspace in ChatGPT. Instead of the model rewriting a whole document in the chat thread each time you asked for a change, your draft opened in its own panel and you edited it in place.

OpenAI introduced it in October 2024 and retired it in May 2026.

The release notes for 28 May 2026 state it plainly: "With this update, canvas will no longer be available in GPT-5.5 Instant or GPT-5.5 Thinking. Writing and coding functionality is now supported directly in chat responses through writing blocks and code blocks."

Paid users kept temporary access through older models, but those models have since been retired on their own schedules, so in practice Canvas is gone for everyone.

What replaced it is not a single feature but two. Writing blocks and code blocks handle the in-conversation editing Canvas was built for. Pages, introduced in late September 2026, handles the collaborative document work that Canvas was increasingly being used for.

Publishing AI-assisted content and not sure it is landing? Get a free SEO audit and see what your pages are actually ranking for.

What Canvas Actually Did

To understand why it was replaced rather than simply deleted, it helps to know what problem it solved.

The problem it was built for

Before Canvas, revising a long piece of text with ChatGPT was awkward. You would paste in a draft, ask for a change to the third paragraph, and get the entire piece back rewritten — often with unrequested changes elsewhere. Finding what had moved meant reading both versions side by side.

Canvas separated the document from the conversation. The draft lived in its own panel, the chat stayed as the place where you gave instructions, and changes were applied to the document rather than regenerated as a fresh reply.

How you opened it

Canvas opened either by typing a slash command in the message box or by selecting it from the tools menu attached to the composer. It could also open automatically when ChatGPT judged that a response was long enough to be worth editing rather than reading.

What you could do inside it

The editing model was built around selection. You highlighted a passage and asked for something specific — rewrite this, expand it, simplify it, cut it down.

Alongside that sat a set of one-click controls. For prose, those covered adjusting the length of a piece, shifting the reading level, changing tone, and a general polish pass for clarity. For code, Canvas offered a review pass, inline comments, bug fixing and porting between languages, and it included a console so snippets could actually be run rather than just inspected.

It also kept version history, so you could step back to an earlier draft after taking a piece in a direction that did not work.

That history was more useful than it sounds. Iterative editing with a language model tends to drift: five rounds of small improvements can quietly produce something flatter than the third version. Being able to go back and compare was a real safeguard against that.

Why people liked it

The appeal was control. Writers could keep a draft stable and change one part of it, which is how editing actually works. Developers got something closer to a scratch editor with a reviewer attached than a chat window that happened to print code.

It also changed the rhythm of working with the model. A chat thread pushes you toward asking for a whole new attempt when something is not right. A document panel pushes you toward fixing the part that is wrong, which usually produces a better result in fewer turns.

Why OpenAI Removed It

OpenAI's stated reason, read against what shipped alongside it, is consolidation rather than retreat.

The functionality moved into the chat response itself. Rather than a separate surface you had to deliberately open, long-form text and code now arrive as structured blocks inside the reply, editable in place. The capability became the default behaviour instead of an opt-in mode.

There is a reasonable argument that this is the better design. Canvas had a discoverability problem — plenty of people used ChatGPT daily for a year without ever opening it. Features that require a slash command to find tend to be used by the minority who read release notes.

The trade-off is that a dedicated panel gives you more room and a clearer separation between instruction and artifact. Some of that is genuinely lost.

The other factor is that Canvas overlapped with things OpenAI shipped later. A separate editing surface, a collaborative document feature and a dedicated coding tool all answer versions of the same question, and maintaining three of them is harder to justify than folding the simplest one into the default experience.

Worth noting too: the removal was tied to models, not to a date on your account. Canvas vanished from the then-current models first and lingered on older ones, which is why some users lost it months before others and why the change felt arbitrary at the time.

Content strategy built on AI drafts still needs the fundamentals in place. Get a free SEO audit to check yours.

What To Use Instead

There are three replacements depending on what you were using Canvas for.

Writing blocks, for prose

If you were using Canvas to draft and revise text, this is the direct successor. Long-form writing now comes back as a writing block within the response, and you edit it there rather than in a separate panel.

The working pattern is the same as it was: be specific about what you want changed and which part you want changed. The instruction "tighten the second section and leave the opening alone" still does more for you than "make this better".

Code blocks, for development work

Code now arrives as a code block in the reply with editing and iteration handled in place. The console that Canvas offered for running snippets is the capability most worth testing against your own workflow, since that was the part of Canvas that went furthest beyond text editing.

If you are doing sustained development rather than occasional snippets, OpenAI's dedicated coding tooling is a better fit than either Canvas or chat ever was. Canvas was always a lightweight option rather than a development environment.

The one habit worth carrying over is scoping your requests to a function or a block rather than a whole file. That was what made Canvas's code review useful, and it works just as well against a code block in a reply.

Pages, for collaborative documents

Pages arrived in late September 2026 and covers the use case Canvas was being stretched to fill — a document that more than one person works on.

Per OpenAI's release notes, you can "turn a conversation into a Page, start from a template, or begin writing directly," then "edit the text yourself or ask ChatGPT to revise it, add charts, or create interactive content." Collaborators can work on the same Page and leave comments, each using their own ChatGPT account.

That is a different proposition from Canvas, which was fundamentally a single-user editing surface. If what you wanted from Canvas was shared drafting, Pages is the thing to look at.

It is also the only one of the three that produces something with a life outside the conversation. A writing block lives in a thread; a Page is a document you can point colleagues at. For anything that needs review or sign-off, that distinction decides which one you want.

How To Move Work Out of Canvas

If you still have documents sitting in Canvas through a legacy model, or you are returning to an account you have not opened in months, treat this as a short cleanup job.

1. Open ChatGPT and check which model your conversations are using. Canvas only ever persisted on the older models, and those have been retired.
2. Work back through the conversations where you used Canvas. The chat threads themselves remain; it is the editing surface that is gone.
3. Copy any draft you still care about into wherever you keep documents — a word processor, a notes app, your CMS. Do not leave the only copy of something inside a retired feature.
4. For anything still actively in progress, start a fresh conversation and paste the draft in. The response will come back as a writing block you can edit.
5. If the document was shared with colleagues, or needs to be, rebuild it as a Page rather than as a chat thread.
6. Update any internal documentation or team instructions that still tell people to open Canvas. This is the step most teams skip and then trip over.

What This Says About Building on AI Features

There is a broader lesson here that is worth more than the feature history.

Canvas lasted about twenty months. It was announced with a blog post, picked up by every AI newsletter, taught in courses, and written into internal workflows at a lot of companies. Then it was folded into something else with a single line in a release note.

That is not a criticism of OpenAI specifically. It is the normal shape of a product category moving this fast. But it does argue for a particular kind of caution: build your processes around the outcome you need, not around the specific button that currently produces it.

A useful test: if a new team member could follow your documented process only by clicking the exact buttons named in it, the process is fragile. If they could follow it by understanding the goal, it survives a redesign.

Practically, that means documenting workflows in terms of what you are trying to achieve — "revise the draft section by section" — rather than the UI path. It means keeping source documents outside the assistant. And it means checking the release notes of any tool you have built a dependency on, because the announcement that a feature is going away is rarely as loud as the announcement that it arrived.

Depending on a platform you do not control is a familiar problem in search too. Get a free SEO audit and see how much of your traffic rests on assumptions worth checking.

Final Thoughts

Canvas was ChatGPT's answer to a real problem: chat is a poor interface for editing a document. For twenty months it was the best way to work on a long draft inside ChatGPT, and the controls it offered — adjust length, shift reading level, polish, review code — were well chosen.

It is no longer available. Writing blocks and code blocks do the in-conversation version of the job, and Pages covers shared documents. If you were a heavy Canvas user, Pages is probably the closer match to how you were working.

The habit worth keeping from the Canvas era is the one that made it useful in the first place: edit in passes, be specific about scope, and keep your own copy of anything you would be upset to lose.

Which model you are on also affects which of these features you see at all, since availability has shifted with each generation. Our guide to which ChatGPT model you should use covers how those differences play out in practice.

And if you are reading a tutorial that tells you to open Canvas, check its date against the May 2026 release note before you spend time hunting for a menu item. A surprising amount of the advice currently ranking for this question was written before the feature was retired and has not been updated since.

Stale content outranking current content is a search problem, not just an AI one. Get a free SEO audit and see whether yours is the version people find.