Feature Request: User-Controlled Priority Sources in ChatGPT Library (For All File Types)

Feature Request: User-Controlled Priority Sources in ChatGPT Library (For All File Types)

The Problem

As ChatGPT becomes increasingly capable of working with documents, code, images, spreadsheets, and external data sources, users are accumulating large amounts of information across ChatGPT Library, Projects, Google Drive, and previous conversations.

However, there is an important feature missing: users cannot explicitly define a global hierarchy of preferred information sources that ChatGPT should consult when performing tasks.

For example, a user may have several versions of the same document, multiple implementations of a program, or different images intended as design references.

ChatGPT may retrieve relevant information, but relevance alone does not determine which source the user actually wants it to prioritize.

A newer file is not necessarily the correct file. A frequently retrieved file is not necessarily the preferred one. And a document generated by ChatGPT is not necessarily more authoritative than a version manually approved by the user.

This creates unnecessary repetition and increases the risk of using outdated or unwanted information.

My Proposal: A Dedicated “Priority Sources” Section in ChatGPT Library

I propose adding a new section called “Priority Sources” to ChatGPT Library, alongside existing sections such as Favorites, Images, and All Files.

Users could place files and other supported content into this section to explicitly indicate that these resources should be consulted before ordinary sources when relevant to a task.

For example, users could configure the following source hierarchy:

  1. Priority Sources — User-designated preferred files and resources.

  2. Google Drive — Connected cloud storage.

  3. Regular ChatGPT Library and previous conversation files — Other accessible resources.

  4. Internet and external sources — Supplementary information when necessary.

Ideally, users should be able to customize this order according to their needs.

This priority system should apply across ChatGPT, rather than being limited to a single conversation or project.

This Should Work With All Supported Content Types

Importantly, this feature should not be limited to Word documents, PDFs, or spreadsheets.

It should support multiple types of information:

1. Programming and source code

A developer could store frequently used code, utility functions, project structures, or preferred implementations in Priority Sources.

When ChatGPT is asked to write or modify related code, it should search these resources first and reuse relevant components whenever appropriate.

2. Images and visual references

A designer could store preferred logos, visual references, brand assets, or example designs.

When ChatGPT is asked to create or edit an image, it should prioritize the relevant visual assets from this section, rather than relying solely on generic references.

3. Documents and templates

A professional could store approved document templates, report formats, or standard operating procedures.

ChatGPT should prioritize these resources when producing similar work.

4. Structured data

Users could store reference datasets, spreadsheets, configuration files, JSON structures, or other frequently used information.

When performing relevant tasks, ChatGPT should consult these resources before retrieving alternative data.

The underlying principle is simple:

Users should be able to tell ChatGPT which resources they prefer without having to repeat those instructions in every conversation.

How It Could Work

The implementation could follow a relatively straightforward process:

  1. When the user submits a task, ChatGPT identifies which resources may be relevant.

  2. ChatGPT searches the Priority Sources section first.

  3. If suitable information is found, ChatGPT uses it as the preferred reference.

  4. If the priority resources are insufficient, ChatGPT continues searching other sources according to the user’s configured hierarchy.

  5. If conflicting information is discovered, ChatGPT identifies the conflict and explains which source was prioritized.

Importantly, the system would not need to load every priority file into the model’s context for every interaction. It would only need to retrieve relevant information from the priority collection.

Users could also be given two configuration options:

  • Global Priority Sources: Applied across all relevant conversations.

  • Project-Specific Priority Sources: Applied only within a particular project.

Priority should not mean blindly trusting everything in a file. The system should still respect the user’s current instructions, security requirements, and applicable constraints.

Why Not Just Use Favorites or Project Instructions?

Favorites are useful for quickly accessing important files, but bookmarking a file is different from explicitly telling ChatGPT to prioritize it as an information source.

Similarly, Project instructions can help establish context, but they are not a complete substitute for a configurable, cross-source priority system.

I believe users need a dedicated and intuitive mechanism for controlling how ChatGPT selects information.

Rather than repeatedly writing instructions such as:

“Use this file first.”

“Do not use the previous version.”

“Follow my existing code.”

“Use my preferred image as the reference.”

Users should be able to establish these preferences directly in the Library.

Expected Benefits

This feature could provide several important benefits:

  • Greater user control: Users determine which information sources should take precedence.

  • Improved consistency: ChatGPT can follow preferred templates, code structures, and reference materials across conversations.

  • Reduced errors: Lower risk of accidentally using outdated or unwanted resources.

  • Less repetitive prompting: Users do not have to repeatedly identify the same preferred sources.

  • Better long-term workflows: ChatGPT becomes more useful for professionals and other users who accumulate substantial personal knowledge libraries.

  • Greater transparency: Users can understand which sources influenced the output and why.

Final Thoughts

ChatGPT is becoming much more than a conversational assistant. For many users, it is evolving into a long-term workspace for programming, research, document creation, design, and knowledge management.

As the amount of accessible information grows, simply being able to retrieve information is no longer enough.

The system also needs to understand which information the user wants to prioritize.

I believe a user-controlled, multimodal Priority Sources system would be a practical and valuable improvement to ChatGPT Library.

It would not necessarily require a new language model. Much of the functionality could potentially be implemented through improvements to file management, retrieval, ranking, and source-selection mechanisms.

My central suggestion is simple:

Let users decide which sources ChatGPT should consult first, instead of requiring them to repeatedly explain their preferences.

I would be interested to hear whether other ChatGPT users have encountered similar problems and whether the OpenAI team has considered implementing configurable source priorities.

Thank you for considering this feature request.