I want to submit formal product feedback regarding the retirement of Canvas.
This change has removed functionality that I used regularly and that the current Writing Blocks feature does not replace.
My primary use case is maintaining long-running projects—especially TTRPG campaigns and extensive worldbuilding—in very large ChatGPT conversations. Canvas was extremely useful because it gave me a persistent document alongside the conversation that both I and ChatGPT could continually reference and update. It functioned as a canonical project document rather than simply another message buried somewhere in the chat history.
With Canvas retired, I now have old conversations containing Canvas documents that I can still see, but the current model may no longer be able to access their contents. That is particularly frustrating because these documents were deliberately created as persistent in-session reference material.
Writing Blocks are **not** functionally equivalent. An editable block embedded in the conversation is fundamentally different from a persistent side-by-side workspace/document that remains part of the ongoing project state.
This is not merely a preference for the old interface. It is a genuine loss of capability. A workflow that previously worked well now either requires repeatedly pasting large amounts of information into the conversation, manually maintaining external documents, or reconstructing project state that already existed inside ChatGPT.
I strongly request that OpenAI either:
-
Restore Canvas or an equivalent persistent side-by-side document workspace;
-
Make existing legacy Canvas documents fully accessible to current models; or
-
Provide a proper migration path that converts old Canvas documents into a persistent format that remains accessible to both the user and the model.
I am a paying ChatGPT user, and Canvas was one of the features that materially increased the value of the product for me. Removing it without an equivalent replacement has made ChatGPT less useful for an established workflow I relied upon.
Please log this specifically as a product regression / feature restoration request and pass it to the relevant product team.