
Episode Summary: In this solo episode, host Cary Weston dives deep into one of the most commonly asked features about ChatGPT: Memory. As summer winds down and routines return, Cary uses this time to unpack what memory actually is in ChatGPT, how it...
Loading summary
A
Hey, it's Cary. In this week's episode, we're Talking memory inside ChatGPT. What is it? Where to see it, how to view it, change it, how to use it, in combination with other things inside ChatGPT, all so you can get the most from it and have the best possible experience. If that's interesting to you, then stick around and I will see you on the other side of the music. Hey there. Welcome to the ChatGPT experiment. This is a podcast designed to help curious folks better understand what ChatGPT is and learn a few nuggets that might help you in your personal or professional life. My name is Kerry Weston. I am your host. And hey, I am glad you're here. This is. What are we going to say? The last episode of the summer? Labor Day's around the corner as I record this and summer is winding down. Remember that parents out there remember the Target ad that came out years ago? The most wonderful time of the year. They play the Christmas music and they show the moms school shopping with the kids. The kids are all miserable and the moms are smiling. It's the most wonderful time of the year. I remember this was. This was stressful when you were a kid because you realized that your freedom was going away. And we're back to routine. So end of the summer. Hope you had a good one. I certainly find myself yearning for the boat, being out at camp, doing nothing, swinging in the hammock, all that stuff, because when we live in, in Maine, you, you know, you live for it because the minute it's gone, it gets. It gets cold. I've got friends in different climates, and they kind of take for granted the warm sun on the beach. I think there's something different about having seasons. You know, you get to appreciate each one. But I think the goal is to live in the moment I love. Fall for me is probably one of my favorites, but I certainly take advantage of the lazy days of summer. So hope you had a good one. Hey. So this week I thought I'd tackle ChatGPT memory, get a lot of questions about memory. And it's one of those things that it's fairly ambiguous, to be honest, because it doesn't seem like there's a lot of fixed rules, and perhaps there isn't. And so I've spent some time interacting with ChatGPT itself to go through some frequent questions and to get some objective answers. So I thought what I'd do today is we'll go through some basic, infrequent questions about memory. And I'LL tie in the quote unquote official answer here with my own experience and kind of wrap in some anecdotes if there's one for you, just to help you get a better understanding of what memory is. Now, it's safe to say that memory is only available in paid versions, not free. So if you're using free, there's a number of features you're not getting. And I've always encouraged you, listen, I've always encouraged you that if you don't know if you want to use paid or free, it's not a long term subscription. So you're not tying into a year or something like that. You can jump in for a month and jump out, right? So for 20 bucks you can play with it for a month and then just jump out and you can always come back. It's very flexible, but there's a lot of features, memory being one of them, that I think makes the experience far better, far more valuable. So let's talk about memory today. We're gonna start. Let's just start with the basics, right? What exactly is memory and how is it different? Well, without memory, every chat is kind of like a blank slate, right? Nothing carries over. Kind of like talking to, let's just say from an analogy point of view, be like talking to a cashier or somebody that doesn't know you, right? Each and every time is an exact brand new fresh conversation versus somebody that you visit on a regular basis. And they know your order, right? They know what you like and maybe they ask you, you want your regular, right? Chatgpt remembers key facts across chats, and it doesn't remember everything. It's kind of remembering a notebook of important details. It's not just a longer thread. It actually has a thought process in and of itself, whether it considers this to be important or not. But it'll grab nuggets along the way and it'll remember key facts across various chats. And that's important to say here as well, is that if you've got a number of different conversations going, it doesn't just remember from one conversation, it will remember from across. So it'll grab big picture things and it'll make a decision as to whether or not it should remember. Because its goal, you know, remember this, its goal is to make itself more valuable to you. So as it's remembering, as it's adding nuggets to its memory, it's trying its best to be helpful. It's trying its best to retain things that will help you in the future. Okay? Which leads to another question, which does memory carry over from one chat to the next? And like I mentioned, yes, that's the point, right? So chat history, all of your conversation, those are all of your previous conversations. You can go back and look. That's not memory. Memory is what ChatGPT itself has deemed worth remembering, right? It saves key points. It does not save the whole conversation. And some people, you know, feel surprising at first, right? It's like, I didn't know it was listening to me all the time. I didn't know it was remembering things all the time. It'll freak some people out, right? But you can stay in control. And we're going to talk about how, okay, we're going to talk about how. What kind of things does ChatGPT decide to remember automatically? Well, if you use it often enough, it's going to remember big picture facts. It's going to try and remember your name, it's going to try to remember your role, it's going to try to remember projects or things that you're working on. Right? It might try to remind, remember things like style, like how you like to receive information. Do you prefer bullets or short answers? Do you prefer thorough? Right. Longer form? It'll try to remember stuff like that. For me, I can tell from firsthand. It remembers ongoing work. So it will remember that I'm working on a podcast. It will remember that I'm writing an article or a book or a guide. It remember that it may talk about styles and cues and tones, right? But it's important to know it doesn't grab everything, right? It just, again, I'm going to come back to this because I think this is the vague term, it's just going to grab what it thinks it's useful, right? And sometimes it does save things that you didn't want it to save. Or like me, I use ChatGPT for things other than my own projects. I use it for client projects. And. And so it may remember something that it thinks is about me, but it's really about somebody else, right? And so another question here is, can I tell it what I want it to remember or not remember? And the answer is yes. You can actually say, as you are going through, just to help it, as you are having a conversation with it, you can say, please remember this or this is important, please remember. And you can also say, forget that. I don't want you to remember this. So if this is in your memory, get rid of it. Right? You can add things on purpose. So I've said once, remember that I host the Chat GPT experiment, right? And I was months ago working on a website with a client. And so I've told it, I don't want you to remember this at all. I don't want you to remember the website project at all. It's not nothing that I want to carry forward. Right. And then you can also edit directly in the memory panel. We're going to talk about that right now. So you're not stuck with what chatgpt thinks is important. You have control over this. You get to curate it. Okay, so how do you check and how do you edit and how do you revise? Okay, so a couple different things I'd like you to do right now, if you're interested in the topic of memory is when you open ChatGPT, say this. Tell me what you know about me, personally and professionally. Okay, that's a. That's just one sentence. I think you'll be surprised if you've been using ChatGPT for any time. Just say, tell me what you know about me personally and professionally, and it's going to give you back what it thinks it knows about you. Okay? And some of these are going to be facts, and some of these are going to be assumptions, and some of these are going to be observations. Now, is that the same as controlling your memory? It is not. But the first thing I want to share with you is that's the exercise that I would start with because I want you to understand how ChatGPT sees you. Okay? The second thing you can do is click on your own account and where you go into settings. And when you click into Settings, you're going to see a tab called Personalization. Again, this is in the paid account. And one of the areas under Personalization is Manage Memories. And if you click that Manage button, you're going to see saved memories. It's going to. You're going to see the nuggets that it is remembering about you, and you can delete any of them that you don't think is important. And then again, the reason I ask you to do that exercise to begin with is tell me what you know about me personally and professionally. You can see why it knows that as it goes through to memory. And so two things can happen here. You can delete memories or facts or whatever is in there that you don't want, and then you can talk to it to say in your memory, I see this, or in the answer of what you know about me, you mentioned this. This is either incorrect or I want to change this or you Forgot this. Like, you can drive it from two different ways. Okay. So you can edit it and view it through managed memories in the personalization of your settings and then ask it to give you that laundry list. And you can kind of go through it and talk and say, I want to talk to you about what you know about me because I want to add, edit and force delete that kind of stuff. So a couple different exercises. And why and why would you want to do that? Well, you want to make sure that it's remembering and helping for what's most important to you, not just what it thinks is most important. Right. So in the trainings I do, when I say when, I tell folks to use that prompt, tell me what you know about me personally, professionally, it's eye opening because there are insinuations and observations. And then if you want to go one step further, if you've been using now, you've got to have a little bit history here. So you know you've got to be using it for a number of months in order for this to be relevant, because the sample size needs to be big enough. Right. But the other thing you can do, if you really want, I've had a number of you write and say you're looking for an accountability partner or a brainstorming partner or a daily planning partner, or a personal improvement partner, is if you've been using ChatGPT long enough, you can say beyond just tell me what you know about me and edit that. You can say, tell me what my five shortcomings are or something like that. And if you want to talk about observations, ChatGPT will bring back what it sees in the work that you're doing, in the conversations you're having and how you use ChatGPT. It's going to formulate some opinions and give you back what it thinks your five shortcomings are. And then you can use that you've heard me share with you before. Tell me more. So don't ever accept everything it gives you as the only thing it knows. Right. So if you say, for instance, I want to work on personal improvement, can you tell me what you think my five biggest shortcomings are? And it will give you what it thinks the answer is in their opinion. Tell me more about number two. Right. And then engage in that conversation. But that's part of its memory. Some of you will see that memory gets full. So what happens when you get that notice that says memory is full? Do I lose old stuff? Well, memory is limited. Yes. It's like a notebook and it's got like a fixed number of pages, right? And when it's near full, you will get a warning. And when that happens, what it considers to be less relevant or even older notes will get trimmed to make space, right? It's more like pruning. So what you'll end up seeing or experiencing is not all at once. Not things aren't just dropping off the shelf, but it's going to start making space by kind of pruning and shrinking some of the older and maybe the stuff that it considers to be not as essential. So part of your exercise, when you take a look at memory and you go through the process of telling, you know, asking it what it knows about you, you could say, this is important. Maybe you could categorize these facts, these are important, and these are less important, and then share with them. If we ever have to make room, these are the things I think we can prune down. But this is the stuff I want to keep. I've had people ask me, if I share a document, can I ask memory to remember the document, right? Can I use documents as part of my memory? So the direct answer is no, it won't store full documents. It will remember that you shared a report or something, but it won't remember the content of the documents, right? So if you want anything to be remembered, you've got to literally tell it what you want it to be remembered. You know, so memory is like context, right? It's the nuances around it. But if you want specific things remembered, like documents and whatnot, it's not going to work for you, right? So it's not going to be your storage center. And we're going to talk in a second about how you can use workarounds. And there's a couple different ways in which tools inside ChatGPT can work well together. So memory is just a piece of the puzzle. It's not the entire toolbox, Right? It's just a tool. So we'll get to that in a second. How does memory work? With custom instructions? And do I need both? So if you haven't done this yet, click on your account and again go back to Settings, and under Custom Instructions is under Personalization. If you haven't done this, go to custom Instructions. And this is where you can tell chatgpt about you. And these are evergreen foundational principles. This is not like memory, where it's going to fade. Whatever you put into custom instructions is absolute, right? So you can tell it what your name is, you can tell it what your preferences are. You can tell it what you do. You can whatever you want to put in there, you can tell it how you prefer to have questions answered. You can style voice, tone and style. You can do all kinds of stuff. This is absolute. Right. And so the ability of using custom instructions for that type of stuff and not having to rely on memory would be good. They work well together and memory is evolving. Notes. Right. So custom instructions are the fixed rules. They're the background, the tone, the format, the information that you give it that never changes. Memory is kind of evolving because as those conversations take place, it's going to learn and see new things that you're doing. So it's going to change pieces of your memory as you work more together, and they do work well together. Memory. Memory will usually win, though, if there's a conflict because it considers memory to be the most recent. However, if you use custom instructions as a evergreen foundational thing, there shouldn't be much in there that will change over time. Right. So the best practice here is to use custom instructions for permanent facts and then the memory will pick up that flexible context as you're using. Right. But I would definitely use both. Right. Use them in tandem. They're both tools. Okay. So custom instructions, Evergreen memory is a flexible kind of remembering things as it goes. Had somebody ask me, what if ChatGPT remembers something I didn't want it to? Am I stuck with it? And the answer is no. Right. You can delete again, going into the settings, going into personalization, and going into memory. You can delete literal things, right. Or through conversation. When you say, show me what you know about me, professionally and personally, you know, you can say, forget this, this is wrong. That's not me. You can have a conversation with it and it'll take care of it with you. So you're in the driver's seat, right? Can you turn memory off? Yes. Even in a paid account, if you don't want any of this to happen in your personalization, you can turn memory off. You do not have to do this. Right. You can go into temporary mode. Somebody asked, when should I rely on memory and when should I paste details in? Well, it's going to be fluid again. It's going to be depending on how you're using ChatGPT, how often you do it, if the memory is correct or incorrect. So I would first go through and I would do that exercise, what do you know about me? And go through that, make sure that you've got that right. And then I would tell it, which is most important and which is flexible, right? We prioritize it so it knows what to focus on and what could be trimmed or pruned later on the road. Use custom instructions for the evergreen, fixed, permanent kind of stuff and then check your memory every once in a while just to see what's happening, right? How often you do that really depends on how often you use ChatGPT, right? And so one of the questions back to the documents, how do I enhance memory? Like, how do I use all of the tools? So memory is going to work on its own and you can enhance it and change it and influence it based on some of the exercises that I've used. Custom instructions is going to be your evergreen. Another layer on top of this, another tool to add to the toolbox is if you use projects, projects allows you to create a singular purpose type entity where you can again create some instructions and upload documents. So you might have a project around your social media postings, right? Or whatever it might be. Think of some specific thing that you're going to do over and over again. You might want to create a project and in the project they do have the ability of uploading documents. Now these documents could be every time I write a social media post for LinkedIn, you know, this is the format, style options I want to use, these are the rules I want to use. And maybe that's a document called LinkedIn rules and you add that every time I use it for an email. These are, this is the rules I want to follow, or these are the, the nuances that I want to embrace. Each of those sets of rules, instructions, attributes, guidelines can be a document inside a project and you can create some instructions inside the project to really help it focus. On top of that, that project is going to tap into your custom instructions that's in your chatgpt and on top of that it's going to borrow from memory. So there are ways in which you can really utilize multiple tools here inside ChatGPT to bring hard details, long format and project and using those documents, the nuances of the flexible memory, asking it what it knows about you and making sure that you're including facts that are true or facts that you want to enhance. Fill in the gaps, delete the obvious, right? Somebody asked, does memory work with custom GPTs? Yes, it does. And custom GPTs are hyper focused on a specific thing, specific thing in your instructions. But it will tag team with your custom instructions and it will tag team with your memory. So keep that in mind. However, sometimes you don't want that. So in your custom GPT, you have your instructions, you can tell it. Pay no attention to memory, pay no attention to custom instructions. Focus only on either the documents or the instructions or whatnot. You can override it, right? So really there's a lot of situations where you want to bring them all together and sometimes you want to separate it and be completely right, completely separate and focused outside of your custom instructions or your memory, whatever it might be. The ability for you to, to understand, I think this is my closing here. The ability for you to understand what ChatGPT knows about you and what you've given it for details is probably one of the more significant elements as we get into paid ChatGPT version enhancements that has come along. Meaning, let me. That was a really sloppy sentence. So let me tell you this. As ChatGPT advances, we're in ChatGPT version 5 now. Memory is really, I think, the biggest area that I've seen be enhanced because it has more capability of storage, more capability of reasoning, more context capabilities. And while I certainly encourage you to go through many of the exercises that I've laid out in earlier episodes, like creating documents about your voice, tone and style and creating documents about rules of engagement and who your audience is and what you're doing. And if you have any formatting styles, get all that out of ChatGPT and stored into documents and then either bring it in as needed in conversations or bring it into projects or bring it into custom GPTs. But memory is helping folks do more on top of that attribute list on top of that document and it can be a really powerful tool. However, like I said, if you are using it for multiple people at work, using it for multiple clients and purposes, please know that ChatGPT is going to remember those nuggets and some of it it may think it's for you and it's not. So cleaning up, helping it refocus, telling it what to remember, what to forget is very, very important. Right? The final nugget here. In one of the earlier episodes I mentioned using ChatGPT as a personal assistant or a brainstorming partner, right? Is I have a weekly routine where the memory feature helps me keep my tasks and my priorities, short term and long term goals on focus. And it knows what the agenda is, it knows what my latest updates on client work is because I'm telling it and it's asking me, it's challenging me every week when I use it. Remember last week I told you I haven't been really good about using it all, but it'll come back and Remember the things that I'm working on, it remembers my goals, it remembers to challenge me, right? And every time I'm done with one of those sessions and I say, give me an updated list as to where we're at, what's the agenda gonna look like next time, that kind of thing, what the status report is or whatever that phrase might be. I actually copy and paste that into a note that I can use next week. Not because ChatGPT is not going to remember, but I want to verify, right? I want to trust, but verify. And so I take the notes. At the conclusion of the session, I copy it and I paste it into a note. So next time I can kind of just make sure that if it's not remembering something, because chances are it will start to forget things. Because I use it so much, I can bring that agenda back in and say, remember this is what we're talking about, and it'll tap into its memory. But I think the ability for this is what's changing for me. The ability for us to identify to ChatGPT what's an important memory and what's a less important memory, that context. So when we start to get into the trimming and pruning phase, it knows which to kind of enable drop off and which is most important, I think is awesome. So I hope that was helpful. Again, like everything else in ChatGPT, there's no gospel, there's no absolutes. You may use it differently than somebody else. I just want you to understand the framework of the flexible memory versus the custom instructions, which is the evergreen fixed principles, versus uploading documents to things like ChatGPT and projects to help expand it. Right. The ability for you to get the most out of this is to better understand it. And understanding memories are the key components. So I hope this was helpful. Again, some of the things I just shared with you will change probably before you know, this episode even goes live. That's how fast things are changing these days. But sharing with you what I know and what I've seen hopefully will go a long way. Getting you familiar with the basics. And then, as always, we adapt as things change. Right? So that's it. That is ChatGPT memory. If you have any questions, you can reach me@chatgptexperiment.com there's some sessions and workshops I've got. Let's see, this fall I'm heading to Washington D.C. for a. For a workshop session, heading out to Vegas for a session, and I've got a couple conversations happening. One in the middle of the country and one down southeast. So it's going to be a busy travel schedule as we get into the fall. But if you have a workshop or a seminar and you're looking for this type of chatgpt kind of hands on interactive training, give me a shout. Find me@chatgptexperiment.com be happy to talk to you about it. One on one trainings as well. And as always the tips and prompts and articles right? As well as the episode archive. So hope things are well with you. I hope your summer was full of memories. Speaking of memory and as we get into the most wonderful time of the year for parents, as we get into the new routine of post Labor Day school here in Maine, I think some schools have already been back. If we're in other places the country. As we get into post Labor Day, we get into the approaching the fall season, right? So thanks again for being with me. Hope this was helpful. As always, your curiosity is the most important. So until we talk soon, be curious. Talk soon. Warning the following ZipRecruiter radio spot you are about to hear is going to be filled with F words when you're hiring.
B
We at ZipRecruiter know you can feel frustrated, forlorn even, like your efforts are futile. And you can spend a fortune trying to find fabulous people, only to get flooded with candidates who are just fine. Fortunately, ZipRecruiter figured out how to fix all that and right now you can try ZipRecruiter for free at ZipRecruiter.com Zip with ZipRecruiter you can forget your frustrations because we find the right people for your roles fast, which is our absolute favorite F word. In fact, four out of five employers who post on ZipRecruiter get a quality candidate within the first day.
A
Fantastic.
B
So whether you need to hire four, 40 or 400 people, get ready to meet first rate talent. Just go to ZipRecruiter.com Zip to try ZipRecruiter for free. Don't forget that's ZipRecruiter.com Zip finally, that's ZipRecruiter.Com Zip Marketing is hard, but I'll.
C
Tell you a little secret. It doesn't have to be. Let me point something out. You're listening to a podcast right now and it's great. You love the host. You seek it out and download it. You listen to it while driving, working out, cooking, even going to the bathroom. Podcasts are a pretty close companion. And this is a podcast ad. Did I get your attention. You can reach great listeners like yourself with podcast advertising from Libsyn Ads. Choose from hundreds of top podcasts offering host endorsements or run a pre produced ad like this one across thousands of shows. To reach your target audience in their favorite podcasts with Libsyn Ads, go to Libsyn ads.com that's L I B S Y N ads.com today.
Host: Cary Weston
Date: August 25, 2025
In this episode, Cary Weston demystifies the “memory” feature in ChatGPT—what it is, how it works, and how you can control and use it to your advantage. He explains the distinction between memory, chat history, and custom instructions, while offering actionable prompts and strategies for curious beginners and frequent users alike. Drawing from both OpenAI’s policies and his own hands-on experience, Cary ensures listeners walk away understanding how to get ChatGPT to better support their personal and professional goals.
Timestamp: 03:45
Timestamp: 08:09
Timestamp: 12:00
Timestamp: 14:20
Timestamp: 17:30
Timestamp: 20:05
Timestamp: 22:03
Timestamp: 24:16
Timestamp: 25:09
Timestamp: 27:24
| Time | Segment | |-----------|---------------------------------------------------------------------------------| | 03:45 | What memory is & why it’s a paid feature | | 05:08 | Distinguishing chat history and memory | | 06:30 | Memory’s process for deciding what to retain | | 08:09 | Specific things ChatGPT tries to remember | | 12:00 | How to tell ChatGPT what to remember or forget | | 14:20 | Key exercises: discovering what memory has stored | | 17:30 | Deleting, editing, and curating your memory | | 18:12 | What happens when memory is full? | | 20:05 | Why memory doesn’t store full documents | | 22:03 | Custom instructions vs. memory: best practices | | 24:16 | Turning memory off and using temporary mode | | 25:09 | Layering: projects, documents, custom GPTs, and memory | | 27:24 | Weekly routines, trust but verify, memory’s role in ongoing planning | | 27:41 | Copy/paste as a verification method |
Cary’s style is warm, practical, and occasionally anecdotal, favoring analogies and step-by-step instructions. He balances technical explanations with relatable metaphors, making complex AI concepts feel accessible and non-threatening—consistent with his signature, “Stay Curious” motto.
| Feature | What it Does | How to Control | |------------------|-----------------------------------------------|---------------------------------------| | Memory | Remembers select facts & context over time | Via settings / direct chat prompts | | Custom Instructions | Stores definitive, permanent user info | Edit under Personalization in settings | | Projects | Houses project-specific docs/instructions | Set up individual projects | | Chat History | Records all chats, does not "remember" facts | Cannot be edited |
Visit chatgptexperiment.com
“Your curiosity is the most important. So until we talk soon, be curious.”
— Cary Weston [29:20]