The Custom GPT Timeline Changed Your Options
If you've built Custom GPTs on your personal ChatGPT account, you're on borrowed time. OpenAI announced in September 2026 that Custom GPTs are being sunset. Personal account creation stopped on August 16, 2026. Enterprise teams lose creation rights on September 25, 2026. Full deprecation hits December 11, 2026. After that date, your Custom GPTs will no longer exist.
This matters because many professionals and creators relied on Custom GPTs to bundle instructions, files, and conversation history into persistent workspaces. The feature felt stable. It was convenient. But OpenAI decided the platform was moving toward enterprise Workspace Agents and plugin-based systems instead, leaving personal users without a clear path forward.
If you're scattered across loose ChatGPT conversations, repeating setup instructions for recurring tasks, or worried about losing work to chat history, you need a replacement. And it exists. ChatGPT Projects launched in December 2024 and remains underused. Unlike Custom GPTs, Projects are available across every account tier—Free, Plus, and Pro—and OpenAI actively maintains them as the stable feature for persistent, organized workflows.
What You Actually Lose Without a Workspace
Most professionals and creators are using ChatGPT in one of two ways: either hunting through old conversations to find relevant context, or recreating the same setup instructions in a new chat every time they need to work on something recurring.
Both approaches waste time and hurt consistency. You lose context between sessions. Each new conversation starts cold. You repeat the same system prompt or role description five times a week. The AI doesn't remember your preferences for tone, output format, or domain knowledge. Output quality drifts because the model isn't anchored to the same instructions every time. If you're working with collaborators, you can't easily share your workspace or let them see your instructions.
A personal workspace solves this. It's a dedicated container for conversations, uploaded files, and persistent instructions that stay consistent across every session. New conversations within that workspace automatically inherit your rules. Files stay organized in one place. Collaborators can see and contribute to the same work. You stop repeating setup and start doing actual work.
ChatGPT Projects: The Core Difference
Projects are not Custom GPTs. They don't publish to a marketplace. They don't have shareable links or a public presence. They're private workspaces within your ChatGPT account.
Here's what they include:
- Conversations. Multiple chats grouped under one project. Each chat inherits the project's instructions and file access.
- Files. Upload PDFs, documents, spreadsheets, images, or code files. Free plans get 5 files per project; Plus gets 25; Pro gets 40. Files persist across all conversations in that project.
- Custom instructions. A Master Prompt that runs at the start of every conversation, setting tone, role, output format, and constraints. No need to repeat it.
- Memory scope. You control whether the model remembers context across projects or keeps each project isolated. Prevents contamination across unrelated work.
- Sharing. Add up to 100 collaborators and give them view or edit access to the entire workspace.
All of this is free on every plan. Projects are available on web and mobile. They integrate with voice input, file uploads, and ChatGPT Work for multi-step production tasks. They're the opposite of deprecated—OpenAI keeps expanding them.
Five Steps to Build Your First Project
1. Name It Like a Specific Outcome, Not a Category
Your naming convention matters more than it seems. Flat names like "Marketing" or "Writing" create vague workspaces that don't tell you what goes in them or when to use them. You'll end up with twelve "General" projects and forget which one handles what.
Instead, use outcome-based names that map to concrete work with a date or client attached. Examples:
- "Client A, Onboarding Docs Rewrite, Q2 2026"
- "SEO Content Production, March 2026"
- "Research: Product Launch Competitive Analysis"
- "Personal Brand Content, LinkedIn + Blog"
- "Legal Document Review and Summarization"
This naming style tells you exactly what the project is for, whether it's active right now, and where its conversations belong. You'll spend less time hunting and more time working.
2. Upload Your Reference Files
Gather the files the project needs: brand guidelines, previous work samples, templates, research documents, competitor analysis, product specs, or client briefs. Upload them all to the project at once. ChatGPT will index them and make them searchable across every conversation.
Don't load unrelated files. Keep each project focused. If a file belongs to two projects, upload it to both. This keeps the AI from chasing context that isn't relevant to the task.
Remember your file limit based on your plan. Free users get 5; Plus users get 25; Pro users get 40. Manage uploads intentionally so you don't hit the cap and have to delete something useful later.
3. Write a Master Prompt (Not a Random Description)
The project description field accepts your Master Prompt—a set of instructions that runs at the start of every conversation in that project. This is where you bake in your voice, role, and output expectations so you don't have to repeat them.
A strong Master Prompt includes:
- Your role. "You are a senior content strategist for a B2B SaaS company." or "You are a research analyst for venture-backed startups."
- Your audience. "Write for professionals with 5+ years of experience in marketing ops who are skeptical of hype."
- Output format. "Structure responses as: summary (one paragraph), key points (3-5 bullets), action steps (numbered list)."
- Tone. "Direct, confident, practical. Avoid filler, generic language, and unverified claims. Prioritize usefulness over comprehensiveness."
- Key constraints. "Never invent data or metrics. Flag assumptions. Cite only the uploaded documents." or "Use short paragraphs, active voice, no jargon."
Start with 150-300 words. Test it with real conversations. Refine it based on whether the outputs match what you expected. A well-tuned Master Prompt is the difference between consistent, reliable work and random, inconsistent responses.
4. Set Memory Boundaries
ChatGPT has memory features that let the model remember facts about you across conversations. In a project workspace, you control whether memory applies project-wide or stays isolated.
If you're working on multiple unrelated projects (e.g., content creation and legal research), isolate memory by project. This prevents the model from confusing a client brief with a competitor analysis or bleeding insights from one project into another.
If you're building a long-term workspace where continuity matters (e.g., an ongoing product launch), enable memory so the model remembers prior decisions and context.
5. Add Your First Collaborators (Or Keep It Private)
Projects can be private (only you) or shared with up to 100 collaborators. If you're working solo, keep it private. If you collaborate with a writer, designer, researcher, or client, add them with view or edit permissions.
Shared projects let team members see all conversations, files, and the Master Prompt in one place. No more forwarding screenshots or context. Everyone works from the same source of truth.
Real Workflows: Three Ways Professionals Use Projects
Content Production: Consistent Voice Across Output
A content creator uploads brand guidelines, previous blog posts, and a style guide to a project. The Master Prompt specifies tone, target audience, and output format (headline, 150-word summary, 5 key points, CTA). Every conversation in that project follows the same structure. The AI remembers the voice. Output is consistent. Editing time drops because the first draft already matches brand standards.
Files stay in one place. The creator doesn't hunt for the brand guide or upload it five times a week. Collaborators can jump in and see the full context immediately.
Research and Analysis: Persistent Knowledge Base
A strategist uploads competitor reports, market research, industry benchmarks, and internal strategy docs to a project. The Master Prompt positions them as a strategic analyst with deep domain knowledge. Every conversation has instant access to that context. They can ask questions, synthesize findings, and generate reports without reloading the same documents or restating the context every time.
The project becomes a growing knowledge base. New research gets added. Conversations stay searchable. Over weeks and months, they build a archive of analysis that they and collaborators can reference instantly.
Client Work: Organized Handoff and Collaboration
An agency uploads client briefs, previous work, approved messaging, and brand assets to a project. They invite the client to view conversations and see work-in-progress. The Master Prompt keeps output aligned with client expectations (tone, format, messaging priorities). Every deliverable reflects the project context.
The client can comment, ask questions, and see the full history of decisions. The agency doesn't send scattered emails or long context-setting messages. Everything is organized in one workspace.
Mistakes That Derail Projects (And How to Skip Them)
Mistake 1: Creating a New Conversation Instead of a New Project. If you're tempted to start a fresh conversation because the current one feels cluttered, pause. That signals you need a new project, not a new chat. New projects have separate files and instructions. New conversations inherit the project's settings but share files and context. Know the difference or you'll end up with messy, duplicative projects.
Mistake 2: Overloading One Project. If a project name is vague (like "General Work"), you'll keep dumping unrelated tasks into it. The Master Prompt will feel generic. Files will compete for the 5-40 upload limit. You'll struggle to find relevant context. One well-scoped project beats ten mixed-purpose ones. Make new projects when the scope changes.
Mistake 3: Ignoring File Limits. Free plans get 5 files. Pro plans get 40. If you upload 6 files to a Free account, you either delete something or lose file access. Track your count. Archive old files. Be intentional about what stays in active projects.
Mistake 4: Treating Projects Like Archived Chats. Projects are workspaces, not history. They're active containers for ongoing work. If a project is done, close it or archive it. Don't keep adding new conversations to an old project just because it's there.
Mistake 5: Not Iterating Your Master Prompt. Your first attempt won't be perfect. Test it for two or three conversations. If outputs aren't what you expected, tweak the instructions. Be specific. Avoid vague language like "be helpful" or "write well." A strong Master Prompt is built through refinement, not guessed at launch.
Expanding Your Workspace: Integration Points
Projects work alone, but they're more powerful when layered with other ChatGPT features.
ChatGPT Work: If you need to run multi-step tasks—like writing a doc, getting feedback, and then refining it—ChatGPT Work lets you chain those steps together within a project. Projects provide the persistent context; Work provides the execution flow.
Scheduled Tasks: Some workflows need to run on a cadence. You can set up a project to run recurring analysis, generate weekly reports, or surface insights on a schedule. This turns a workspace into a production system.
Voice Input and Mobile: Projects work on mobile and support voice input. If you're on the go and need to add context or start a conversation, you can use voice to interact with your project without typing.
Memory Across Sessions: If you enable memory in a project, the model builds a persistent understanding of your preferences, past decisions, and context over time. This is powerful for long-running projects but risky if you want isolation between unrelated work.
Start with the core five steps above. Add these features only when the core workflow is stable and you see a clear use case.
The Difference Between Now and Later
Custom GPTs are gone by December 2026. If you're still using them, you're weeks away from losing access. If you haven't built one yet, you can't—personal account creation ended in August 2026.
Projects are the alternative, and they're actually better for most professionals. They're available on every plan. They don't get deprecated. They work on mobile. They integrate with emerging ChatGPT features like Work and scheduled tasks. They're actively maintained. They solve the real problem: organizing persistent work without losing context or repeating setup.
The goal is simple. Stop hunting through chat history. Stop repeating instructions. Stop watching output quality drift because the model isn't anchored to consistent rules. Build a workspace that persists, that your collaborators can access, and that actually survives the next platform change.
Your Next Step
Pick one recurring task you do in ChatGPT at least twice a week—writing, research, analysis, or content production. Create a new project for it using the five steps above. Name it with an outcome and date. Upload your reference files. Write a Master Prompt that captures your role and output expectations. Run three conversations. Refine the prompt based on what you see. That's your baseline.
If you're evaluating whether ChatGPT Projects fit into a larger AI workflow—or comparing it to other options—use the AI Stack Recommender to see how ChatGPT fits with other tools based on your budget, team size, and technical needs.
Once your project is running smoothly, read How to Use AI to layer in systematic review and measurement so you can quantify the time and consistency gains. Then expand to other recurring workflows using the same pattern.
This is the durable path forward. It works today. It'll work in 2027. It doesn't disappear when OpenAI changes priorities.