ChatGPT memory for documents is now rolling out to free users, letting the assistant recall information from uploaded files across multiple conversations instead of treating each chat as isolated. OpenAI began rolling out memory improvements for free users on June 3, 2025, though the feature set varies significantly between free and paid tiers.
Key Takeaways
- Free users get lightweight ChatGPT memory for documents with short-term continuity across conversations.
- Plus and Pro users receive longer-term understanding of user context and preferences.
- Uploaded files are saved to a Library for later reuse, though Library access is restricted to Plus, Pro, and Business users in most regions.
- Memory can be disabled in Settings; users can also switch to Temporary Chat mode to exclude conversations from memory.
- Individual file uploads are capped at 512MB, with text documents limited to 2 million tokens.
What ChatGPT memory for documents actually does
The feature lets ChatGPT reference previously uploaded documents, spreadsheets, and images days or weeks after you first shared them. Instead of re-uploading a file or restating context, you can ask ChatGPT to recall information from something you uploaded in an earlier conversation. If memory is enabled, ChatGPT may use files and chats to remember useful details between sessions. This shifts ChatGPT from a stateless chat tool into something closer to a persistent workspace.
The difference between free and paid users matters. Free users get what OpenAI calls a lightweight version with short-term continuity across conversations. Plus and Pro subscribers receive longer-term understanding, meaning ChatGPT builds a richer profile of your preferences and context over time. Neither tier offers permanent, indexed recall across all conversations — the system prioritizes recent interactions and explicitly saved memories.
How to use and control ChatGPT memory for documents
You do not need to do anything special to enable the feature. If you are on a compatible plan, memory is on by default. To access previously uploaded files, look for the Library in the left-hand sidebar. You can search and filter files by type: images, documents, spreadsheets, presentations, and PDFs. Files uploaded to a GPT or ChatGPT conversation are automatically saved to the Library so you can reuse them later.
If you want to disable memory, go to Settings and turn off references to saved memories or chat history. You can also use Temporary Chat mode for conversations that do not use or update memory at all. This is useful if you are discussing sensitive information or testing ChatGPT behavior without influencing future interactions.
What you need to know about file limits and regional access
OpenAI imposes strict file size limits. Individual uploads are capped at 512MB per file, and text documents cannot exceed 2 million tokens. Spreadsheets and CSVs are limited to approximately 50MB depending on row size. These constraints mean large datasets or lengthy documents may need to be split or summarized before upload.
Library access is not universal. The Library feature is available only to ChatGPT Plus, Pro, and Business users, and only outside the European Economic Area, Switzerland, and the United Kingdom. Free users in those regions may not have full access to file storage and retrieval. Additionally, the Library is web-only, so you cannot access saved files through the mobile app.
ChatGPT memory for documents vs. older behavior
Previously, ChatGPT treated each conversation as independent. If you uploaded a spreadsheet in one chat and wanted to reference it again later, you had to re-upload it or paste the data again. ChatGPT memory for documents eliminates this friction. The system now understands which files matter to you and can pull relevant details from them without you having to re-introduce the context. This is not the same as permanent indexing — the system still prioritizes recent interactions — but it is a meaningful step toward making ChatGPT a practical tool for managing personal and professional files.
The feature also differs from simple chat history. Chat history lets ChatGPT remember what you said in a previous conversation. ChatGPT memory for documents goes further by letting it recall the actual content of files you uploaded, not just the text you typed. This distinction matters for workflows where files are the primary source of information.
Should you worry about privacy?
OpenAI says users can control memory in Settings and can turn off references to saved memories or chat history. If you disable memory, ChatGPT will not use files or chats to build an understanding of your preferences. However, files you upload may still be used to help improve ChatGPT unless you explicitly opt out. Check your Settings and Privacy controls to see what data OpenAI retains and how it uses uploaded files.
Is ChatGPT memory for documents available on mobile?
The Library is web-only, so you cannot browse or search saved files on the mobile app. You can still upload files through the mobile app, but accessing them requires switching to the web version. This is a notable limitation for users who primarily interact with ChatGPT on their phones.
Can I delete files from ChatGPT memory?
Yes. You can manage files in the Library by removing them if you no longer want ChatGPT to reference them. Deleting a file from the Library removes it from memory, though OpenAI may retain a copy for other purposes depending on your account settings and regional privacy laws.
ChatGPT memory for documents is a practical feature that reduces friction for anyone using the tool to organize or reference files. Free users gain real value here, though paid tiers unlock longer-term context and richer continuity. If you regularly upload documents to ChatGPT, enabling memory transforms the tool from a disposable chat into something you might actually rely on for work or personal projects. The file size limits and regional restrictions are real constraints, but for typical use cases — PDFs, spreadsheets, notes — they are unlikely to matter.
Edited by the All Things Geek team.
Source: TechRadar
