Claude Projects vs ChatGPT Memory: Stop Repeating Yourself to AI

I spent my first six months with ChatGPT explaining the same context in every single conversation. "I'm a productivity consultant. I work mostly with marketing teams. I prefer British English. I like concise responses." Copy, paste, begin.

It felt absurd. Here was this remarkably intelligent system that could write code, analyse data, and draft compelling content—but it couldn't remember that we'd spoken yesterday. Every session was a first date, and I was exhausted from the introductions.

That's changed. Both Claude and ChatGPT now offer ways to maintain persistent context, though they approach the problem very differently. Understanding these differences will save you hours of repetitive setup—and transform how you work with AI.

The Groundhog Day Problem

Before we look at solutions, it's worth understanding why this matters so much.

When you work with a colleague, they accumulate knowledge about you over time. They learn your preferences, your work style, your ongoing projects. You don't re-explain your job description every morning. The relationship builds on itself.

AI assistants, until recently, couldn't do this. Each conversation existed in isolation. The brilliant analysis you got yesterday? The AI doesn't remember it. The detailed context you provided about your client project? Gone. You're perpetually starting from scratch, which means you're perpetually losing time to setup and context-building that should only happen once.

"The most powerful AI feature isn't better reasoning or longer context windows—it's simply remembering who you are."

Two Philosophies of Memory

Claude and ChatGPT have taken fundamentally different approaches to solving this problem, and the difference reflects deeper philosophies about how AI should relate to its users.

Claude's approach is project-based. You create discrete workspaces, upload relevant documents, and write custom instructions. The context is explicit, organised, and under your control. It's like setting up a well-organised filing cabinet before you start working.

ChatGPT's approach is relationship-based. The AI learns about you organically through your conversations, building a picture of who you are over time. It's more like how a human assistant would gradually come to understand your preferences through observation and interaction.

Neither approach is universally better. They excel at different things, and understanding when to use each will make you dramatically more effective.

Claude Projects: The Organised Workspace

I first set up a Claude Project for a client rebrand I was consulting on. The project involved brand guidelines, competitor analysis, stakeholder interview transcripts, and previous creative work—easily fifty pages of material I'd been manually referencing in every conversation.

Setting up the project took twenty minutes. I uploaded the documents, wrote custom instructions explaining the project context and my role, and started a conversation. The difference was immediate: Claude could reference specific details from the brand guidelines, recall insights from the stakeholder interviews, and maintain consistency across our discussions.

Anatomy of an Effective Claude Project

1

Project Name

Clear and specific: "Acme Corp Website Redesign" not "Client Work"

2

Uploaded Documents

Brand guidelines, briefs, reference materials, codebases, previous work

3

Custom Instructions

Your role, the project context, tone requirements, constraints to remember

4

Organised Conversations

Multiple chats within the project, all sharing the same context

Where Claude Projects Excel

Projects shine when you have substantial reference material and need the AI to stay grounded in specific documents. I've found them particularly valuable for:

Client work is the obvious use case. Upload the brief, brand guidelines, previous deliverables, and any relevant research. Every conversation starts with full context, and you can be confident recommendations align with the actual source material—not hallucinated details.

Coding projects benefit enormously. Upload your codebase, and Claude can reference your actual implementation when suggesting changes. No more generic advice that ignores your existing architecture.

Research becomes dramatically more useful. Gather your papers, notes, and sources in one project. You can query across everything, ask for connections between sources, and build analysis that's grounded in your actual materials rather than general knowledge.

ChatGPT Memory: The Learning Relationship

ChatGPT's memory feature works almost invisibly. You don't set it up—it happens. As you chat, GPT notices information worth remembering and stores it for future conversations. Your role, your preferences, your ongoing projects, your communication style.

The first time I experienced this working well, I was startled. I'd asked for help with a presentation, and ChatGPT opened with "Since you mentioned you're presenting to the marketing team at [client name]..." I hadn't mentioned the client in this conversation. It remembered from three weeks earlier.

Active Memory Management

While memory accumulates automatically, you can also actively shape it. This took me a while to realise, and it changed how useful the feature became.

You can also review everything ChatGPT has remembered about you. Navigate to Settings → Personalization → Memory and you'll see a list of stored facts. It's worth reviewing periodically—I found mine included outdated project references and a few facts I'd rather it didn't retain.

Where ChatGPT Memory Excels

Memory works best for ongoing personalisation that should apply everywhere, regardless of topic:

Personal preferences are the clearest win. Your coding language of choice, your writing style, whether you want detailed explanations or quick answers—these preferences carry across every conversation without you mentioning them.

Work context that stays relatively stable benefits from memory. Your role, your company, your team structure. Things that would be tedious to re-explain but don't change week to week.

Communication patterns get better over time. If you consistently edit outputs in certain ways, GPT learns and adjusts. The relationship genuinely improves with use.

Choosing Your Approach

After using both systems extensively, I've developed clear rules for when to reach for each:

The honest answer, though, is that I use both. Claude Projects for deep work on specific initiatives. ChatGPT for general assistance throughout the day. They complement each other well.

Setting Up for Success

If you're starting from scratch, here's the approach I'd recommend:

For ChatGPT Memory

Start by explicitly telling it the core facts about yourself and your work. Don't wait for it to figure things out organically—seed the relationship with the essentials:

Then let it learn naturally from your interactions. Check in on stored memories every few weeks and clean up anything outdated.

For Claude Projects

Start with your most document-heavy recurring task. For many people, that's a specific client engagement or an ongoing internal project. Create the project, upload the core documents, and write thorough custom instructions.

The custom instructions matter more than most people realise. Don't just describe the project—describe how you want Claude to approach it:

Once you've experienced the difference with one project, you'll naturally see opportunities to create others.

The Bigger Picture

Persistent context might seem like a minor convenience feature, but I think it represents something more significant in how we'll work with AI going forward.

When an AI remembers you—your work, your preferences, your ongoing projects—the relationship shifts. You stop thinking of it as a tool you use occasionally and start treating it as a collaborator that's genuinely part of your workflow. The interactions become richer because you're not constantly rebuilding foundation.

We're still early in this evolution. Current memory and project features are useful but limited. Future versions will remember more, understand context better, and integrate more seamlessly into how we work. But you don't need to wait for perfect—the tools available today are already transformative if you invest a few minutes in setting them up.

Stop repeating yourself. Build the context once. Let the AI do what it's good at.