ChatGPT Projects Broken? Troubleshoot Files, Chats, and Memory
What to check when ChatGPT Projects is not loading, files are missing or chats are ignoring the project context. Step-by-step fixes and how to protect your work.
Updated 5 min read
ChatGPT Projects can fail in a few distinct ways: the project does not load, uploaded files do not appear, the assistant ignores the custom instructions, or a chat behaves as though it is not inside a project at all. Most of these have a straightforward fix. Work through the checks below in order before assuming the problem is a platform bug.
If work must continue in another chat or another tool while you troubleshoot, having your work context and next steps recorded somewhere outside ChatGPT matters. Tools like fenbs keep that record so you can hand work to any connected assistant without losing the context.
Project not loading or showing an error
Reload the page or the app. ChatGPT occasionally fails to load a project due to a temporary network issue or a session problem that a fresh load resolves. If you are on the web app, try a hard refresh. If you are on mobile, close and reopen the app.
If the project still does not appear, check OpenAI's status page (opens in a new tab) to see whether there is an ongoing incident. If the platform is down, the fix is to wait. If the status page reports no issues, try a different browser or device to rule out a local problem.
If you were recently on a different account in the same browser, signing out and back in can clear session issues that cause projects to show as empty or inaccessible.
Files not appearing in a project
If a file you uploaded is not visible in the project, first check whether the upload completed. Large files can time out silently. Try uploading the file again and watch for a confirmation indicator before navigating away.
Also check the file type. ChatGPT Projects support a specific set of file formats, and an unsupported format may appear to upload without actually being added. See OpenAI's help center (opens in a new tab) for the current list of supported formats.
If you have uploaded many files, check whether you have reached the project's file count or storage quota. Files beyond the limit may not be added even when no error is shown. The upload limits article covers what to check.
Chats not following custom instructions
If the assistant is ignoring the project's custom instructions, open the project settings and check that the instructions are saved correctly. An empty field or a very long instruction that was truncated can cause the assistant to behave as though no instructions are set.
Also check that the chat is actually inside the project. It is easy to start a new conversation from the main chat screen rather than from inside the project. A chat outside the project does not use the project files or instructions. Look for the project name near the top of the chat interface to confirm.
If the instructions are saved and the chat is inside the project, try rephrasing. Very long or complex instructions can produce inconsistent results. Shorter, more direct instructions are more reliably followed.
Keep the context outside the project
If a ChatGPT project is unavailable or broken, fenbs still has the project context, the decisions and the next step. Any connected assistant can read the context and continue the work. Keeping the decision record in fenbs means a platform outage costs you a project session, not the knowledge that the project depends on.
Create a free accountFree for one person · No credit card · Works with Claude, ChatGPT, Cursor and more
Memory not persisting between chats
Each chat inside a project is independent. The project files and custom instructions are shared, but the conversation history is not. If you open a new chat inside the project and the assistant does not know what you discussed in a previous chat, that is by design, not a bug.
To carry specific information forward, paste a brief summary into the new chat or record the key points in a shared note outside ChatGPT. The most useful things to carry forward are decisions made, open questions and the exact next step.
When multiple people work on the same project, each person's chats are independent of the others' as well. There is no shared conversation history between team members. If a colleague's session covered important ground, they need to copy the relevant points into a shared document, a project file or a tool like fenbs that all connected assistants can read.
When to start a new project
If a project is consistently slow, producing errors or behaving unpredictably, starting a fresh project and moving your files is sometimes faster than continuing to debug. Before doing this, export or copy anything you want to keep from the existing project's chats, since chat history does not transfer when you recreate a project.
See ChatGPT projects best practices for guidance on keeping projects lean and well-organized to reduce the chance of hitting these issues in the first place. A clean project with focused files and short instructions is easier to recover when something goes wrong.
How to do this with fenbs
- Create a free account and add a project for your ongoing work.
- Record your project context, decisions and current tasks in fenbs rather than only in ChatGPT project instructions.
- Connect your AI assistant from Integrations. If ChatGPT is unavailable, connect Claude or another supported assistant to the same fenbs context.
- Before ending each session, ask the assistant to save progress in fenbs so the next session has an accurate starting point.
- If you need to move to a new project or a different tool, the fenbs record survives the move intact.