SUBSCRIBE | ADVERTISE

ChatGPT Projects 101

Why you should be using Projects + advanced tips

Was this email forwarded to you? Sign up here.

TLDR;

• Projects gets you higher quality responses with an AI that has more targeted context. It also saves you time by organising your files and eliminating the need to re-explain yourself.

• Projects work way better with ChatGPT in the browser (Desktop app lacks some key features).

• Pro tips: Project-contained, custom instructions for citing your chats & files, Deep Research inside projects (restrict to specific websites when possible).

• I’m using Projects for both short-term projects and ongoing work.

• Projects does have some important limitations, most importantly a bit unreliable retrieval.

Howdy wizards,

The Projects feature in ChatGPT were launched back in December 2024 (a millenia ago in AI time). However, I’ve noticed many people don’t even know about this feature, or know about it but treat it merely as a way to group your chats so they can find them more easily.

Projects aren’t just great at organising your work, they can also supercharge it. Especially with some recent updates which have gone under the radar, even for me.

I’ll give you a rundown here on Projects and my top tips for using them.

Why you should be using Projects

ChatGPT Projects is a place where you can gather chats, instructions and up to 40 files about a specific topic in one place. You can create a new project from your ChatGPT sidebar, and every chat you start from within that project becomes part of it.

Using Projects eliminates the need for you to explain the context about a topic to ChatGPT over and over. ChatGPT gets context from all the chats within that project. That helps you stay organised and saves you time and cognitive resources. It also helps you get better quality responses (more on that in a bit).

Why not just use a single chat instead? Because as your conversations get longer, AI’s responses gradually get slower, more confused, less reliable, sometimes hallucinations. Not ideal.

“But what about ChatGPT’s standard, cross-chat memory?” you might ask. “Isn’t that enough?” ChatGPT does remember high-level details from your previous conversations, but you simply can’t rely on the built-in memory feature to keep all meaningful aspects of any particular conversation in your history of what I assume is hundreds of convos.

Pro tip: It’s easy to add existing chats into a project (that you originally did outside of that project) by right-clicking it and selecting “Add to project →”.

IN PARTNERSHIP WITH SHUTTERSTOCK

Training cutting edge AI? Unlock the data advantage today.

If you’re building or fine-tuning generative AI models, this guide is your shortcut to smarter AI model training. Learn how Shutterstock’s multimodal datasets—grounded in measurable user behavior—can help you reduce legal risk, boost creative diversity, and improve model reliability.

Inside, you’ll uncover why scraped data and aesthetic proxies often fall short—and how to use clustering methods and semantic evaluation to refine your dataset and your outputs. Designed for AI leaders, product teams, and ML engineers, this guide walks through how to identify refinement-worthy data, align with generative preferences, and validate progress with confidence.

Whether you're optimizing alignment, output quality, or time-to-value, this playbook gives you a data advantage. Download the guide and train your models with data built for performance.

Download the guide

Contain your context within Projects

For a long time, Projects would inherit the memory from all your chats (including those outside of the project). Recently, OpenAI added the ability to contain the project’s memory to that project. When only chats and files within that project are available as memory it means less information—and more targeted information—will go into the LLM’s context window.

I recommend enabling “project-only memory” when you create a new project. The sandboxed environment helps AI stay on track and give you the best possible results.

Pro tip: For max control of your context, use deliberate chat and file names and add these instructions to your project: “ALWAYS cite sources from this Project when you use them in your answers. Use chat/file title + date (dd-mm-yyyy) when you do.”

Deep research inside Projects

Projects + Deep Research are a match made in heaven because it grounds the research (which is based on public web results) with your instructions, chats and files.

Just enter your query, answer a couple of follow-up questions from ChatGPT to narrow down the topic further, then go grab some coffee while it collects information and links from across the web.

Pro tip: when using Deep Research—whenever it makes sense—give it a specific list of websites that are credible and relevant to your project, and ask it to focus its research there.

Where Projects fall short

The main things I currently don’t like with Projects are:

• No good solution to sync with your knowledge base. OpenAI has synced connectors for Google Drive and GitHub where ChatGPT can access your data in real-time, but it only works for Pro/Enterprise/Business tiers and Google Workspace (the business version of Drive) and only for the Drive as a whole—not project-specific folders. My bet is this type of sync-functionality with popular storage/file hosting solutions will soon become better and more widely available in all the leading chatbots.

• ChatGPT unfortunately doesn’t always scan all the relevant information stored in your project when answering. If you want to be absolutely sure certain files or chats are referenced, it’s best to ask the AI to review those files specifically in your prompt.

• There’s huge potential in making the work in Projects more agentic. I want a Cursor-like agent to live inside my project! It shoul search through all the relevant files, and have the ability to take on my docs all by itself. (I saw Notion recently released a more agentic experience, exactly in that direction.) Hopefully this is coming. It will also be interesting to see what happens when Google launches Projects for Gemini (which is on their agenda) as they’ll probably be able to offer the tightest integration with their own apps.

How I’m using Projects

I’ve been using a Projects in lots of different ways so far and I find it, despite the name, not only good for temporary endeavours but also for ongoing pieces of work and research. Here’s a couple of examples where using Projects made a big difference for me.

Example 1: Product positioning study

• What I did: This year I did a product positioning for a B2B software product; a weeks-long endeavour consisting of customer research, plus market, competition and brand analysis. I created a Project in ChatGPT with all the customer interview transcripts, survey data, market research findings, detailed competitor and feature comparisons, and more.

• How it helped: The Project came in handy during the research stages to explore the raw data (e.g. extracting insights from customer interviews, surveys and reviews), do deep research on very specific topics (e.g. “Do deep research to gather all posts from [this] forum that revolve around problems X, Y, Z; with dates and direct links”). During the last stages, the positioning development itself, I used the Project as an advisor to help me brainstorm, validate and refine my ideas for a value proposition, differentiators, messaging hierarchy and proof points—with responses directly grounded in all the research I had done.

Example 2: Building a media kit

• What I did: I’m currently in the process of building a new media kit for this newsletter; a document to help potential sponsors decide if this publication is a good fit for them. I have a Project consisting of previous newsletter issues, all e-mail replies I’ve ever received from readers, survey data from hundreds of you about your AI-related challenges and goals, my ideas for where I want to take the newsletter, branding elements, and more.

• How it helped: It makes analysing my audience and my content much easier, helping me decide what to include in the media kit and how to word it. Now that I have this Project set up, I’ll keep adding to it and start using it more broadly as a sounding board for newsletter content (e.g. finding gaps between my readers’ challenges and my current content to get ideas for future topics).

By the way, there's also a very similar feature (also named Projects) inside Claude, and potentially coming soon for Gemini. While this was all about Projects in ChatGPT, you should be able to apply most of the learnings to other platforms as well.

THAT’S ALL FOR THIS WEEK

Affiliate disclosure: To cover the cost of my email software and the time I spend writing this newsletter, I sometimes link to products and other newsletters. Please assume these are affiliate links. If you choose to subscribe to a newsletter or buy a product through any of my links then THANK YOU – it will make it possible for me to continue to do this.