ChatGPT's new memory builds a profile of you automatically—and OpenAI admits you can't see it all

I'm excited about ChatGPT's memory upgrade - but I'm quickly seeing a downside

Follow ZDNET: Add us as a preferred source on Google. ZDNET's key takeaways * ChatGPT's memory now builds a profile from past chats. * Old or irrelevant details can distort future AI answers. * Turning memory off may not fully erase what ChatGPT knows. According to a blog post released recently, OpenAI seems quite proud of the "improvements" it's made to managing user memories in the chatbot. I'm not sure I like them. In fact, I know I don't. I'm just not sure whether turning the improvements off will make things better or worse. Also: ChatGPT vs. Gemini's AI image generation: A single prompt tweak makes a big difference Memories, for the purpose of this discussion, are details you share with ChatGPT. Introduced in 2024, memories were basically a list of facts the AI could reference. Today, they have expanded considerably to include your entire chat history, explicit instructions, personal constraints, and even implicit preferences the AI derives from past behavior and casual remarks. (Disclosure: Ziff Davis, ZDNET's parent company, filed an April 2025 lawsuit against OpenAI, alleging it infringed Ziff Davis copyrights in training and operating its AI systems.) This article has three sections. First, I'll go over the technical information from OpenAI's recent blog post about ChatGPT's improved memory capabilities. Second, I'll show you the interface elements you can use to tweak ChatGPT's memories. Finally, I'll wrap up with why this feature freaks me out and why it might worry you, too. Let's dig in. Dream a little dream of me Before 2024, ChatGPT didn't have memories of any kind. Each chat session lived on its own, and whatever you told it was unavailable to any other session. Also: Compare new models with our AI Model Release Tracker Then, in 2024, OpenAI introduced memories. In the 2024 context, memories were basically a list of facts. ChatGPT still has this feature, and much of it is absolutely useless. Here are three items still in my ChatGPT memory: * Has two Google Workspace accounts, one for A.com (used for mail and 57 TB of backups) and another for B.com. User wants to move A.com email accounts and routing to the B.com Workspace but retain the A.com Workspace for storage, possibly under a different domain. * Has a global list called 'veg_positions' in their Scratch project to track sprite positions, especially to prevent overlaps among multiple broccoli clones. * Has installed the Kasa smart plug. These items are still in my ChatGPT memory, even though they were only relevant for the one chat session they were discussed in, and even though those sessions were a few years back. This searchable list of saved facts misses implicit context and subtext. As you can see, saved memories went stale fast. Unless curated by the user by hand, there was no mechanism to know when the information became no longer useful. Also: Anthropic launches Opus 4.8, with honesty as its killer feature Then, in 2025, OpenAI began work on what it calls dreaming. For humans, dreaming helps us process our emotions, consolidate and encode memories, simulate potential threats, and reinforce social and emotional intelligence. A Scientific American article on the science behind dreaming says, "Dreams help regulate traffic on that fragile bridge which connects our experiences with our emotions and memories." My dreams usually involve villagers with pitchforks and torches chasing me, or being trapped in caves with flaming skulls and the occasional IBM mainframe. In any case, in 2025, OpenAI's dreaming capability meant the model could reference your chat history in the background, without being explicitly told to look something up or remember. It started to curate memories automatically. ChatGPT's memory structure consisted of saved memories (the 2024 mechanism) plus Dreaming V0. Also: The 5 myths of the agentic coding apocalypse Now, in 2026, ChatGPT's Dreaming feature is at V3. Saved memories still exist, but they're either replaced or augmented with dreaming-based memories. For example, I just asked ChatGPT, "Do I have any experience with Kasa?" It replied, "Yes. You've used or discussed Kasa KP125M Matter smart plugs, specifically for smart home energy monitoring." As you saw above, that model number is not in the saved memory. But it's also completely wrong. The second part of the answer to my "experience with Kasa" question was this: "You later moved that monitoring setup into Home Assistant, so your Kasa experience appears to be tied to power/energy tracking, not just basic on/off plug control." Yeah, not so much. I have never installed Home Assistant. The Kasa plug is currently sitting in my gear bin along with a bunch of other hardware. When I reached out to OpenAI via their PR firm about my concerns, a company representative told ZDNET, "What you're seeing is a new high-level memory summary, rather than a complete inventory of facts ChatGPT may remember. It's meant to make the overall picture easier to review and correct, but it won't necessarily show every detail, such as your tech stack, even when that context can still be used in a relevant conversation." Today's Dreaming V3 doesn't just scan your chat history in the background. It performs data synthesis, effectively composing a dossier about you (that's not always accurate). According to OpenAI, V3 is capable of carrying forward complex context information and tracking dense, multi-session, multilayered long-running projects. According to the company, factual task recall success jumped from 41% in 2024 to 82% in 2026. The ability to stay correct over time went from a 9% taskrate in 2024 to 75% in 2026. Preference adherence went from 31% in 2024 to 71% in 2026. Also: Your Claude agents can 'dream' now - how Anthropic's new feature works How does all of this autonomous background processing scale? That's a big part of the breakthrough that makes Dream V3 possible. OpenAI has slashed the compute cost by 5X for this level of ongoing analysis. The direct result of this efficiency makes it practical for OpenAI to offer the feature for mass access. The AI is constantly revising its internal mental state in the background, looking at timestamps attached to informational nuggets, and effectively experiencing passages of time alongside you (in theory). The Dream V3 feature is available now to Plus and Pro tier subscribers. The feature will be rolling out to all users (including free users) over the coming weeks. A trip down Memory Lane I found the new memory features in the browser version of ChatGPT. At the time of this writing, the MacOS app still has the old memory interface. To access memory, go to Settings, then Personalization. Scroll down to the Memory section. You can disable the memory capability (shown at 1), but only partially. If you turn off Enable memory, your already-stored memories will not be removed, nor will any of your chat history. ChatGPT won't do dream-based memory consolidation, but the data will be there. Also: Anthropic's Mythos is evolving faster than expected, reports AI safety agency If you want to delete your memories, go to the saved memories interface (at 2) and delete saved memories. Even then, all your memories won't be deleted. You'll have to delete the actual chat itself to purge that information from the AI's consciousness. There's also a weird gotcha. In ChatGPT's memory FAQ, OpenAI says this: Turning off Memory/Personalization does not disable safety features that may use limited, safety-relevant context in rare, high-risk situations to help ChatGPT respond more safely. You can read OpenAI's description of its safety protocols. However, the basic idea is that if a conversation implies a safety issue, ChatGPT won't delete that information and will use it to help the user de-escalate a situation. It doesn't appear like the AI reports that information to anyone, but it's there (and somewhat unclear). Finally, you can hit Manage (at 3) to make changes to the consolidated profile ChatGPT has developed about you. Also: I tried Google Drive's new AI cleanup tool to fix 14 years of storage clutter Let's tap into saved memories first (at 2). Here's an excerpt of mine. As you can see, they're highly specific (and quite irrelevant or wrong). For example, I'm not running Mint Linux on my Mac Studio, but I was exploring the idea about a year ago. The new feature is under the Manage button. Because I don't want to share ChatGPT's interpretation of my personality and preferences, I'm using a screenshot OpenAI provided in its blog. As you can see, it's a narrative about interests and preferences. You can select aspects and mark "Don't mention this again" to effectively forget that item. You can also add comments to personalize how ChatGPT groks you. Thanks for the memory There are scenarios where solid AI memory capability is mission-critical. I've constructed a very precise set of memory instructions that govern how Claude Code manages my Apple coding projects. I have another memory infrastructure running on my OpenClaw home server that allows it to retain context and instructions between tasks. Also: I used Claude Code to vibe code a Mac app in 8 hours But I'm not at all comfortable with this new memory mechanism in ChatGPT. I showed above how the feature retains wildly out-of-date information and then uses this information in its responses. Even Dreaming V3, which supposedly changes with you over time, can be wrong. I showed you the example where it claims I'm running Home Assistant, and I've never even downloaded it. It filters the entire worldview through a personal lens that's somehow derived from discussions and preferences. This is a behavior I not only dislike but also find problematic as heck. ChatGPT doesn't really know who I am because it derives its assumptions based only on what I show it. It's like people thinking they know me solely from my social media posts, or my great Aunt Sally, who still thinks of me as a third grader even though I'm a successful professional with an advanced degree, or an old buddy who still thinks I eat only junk food and am incapable of appreciating nice things because I once went through a low-budget convenience phase back in the 1990s. Also: This exec offers 4 ways to be a successful innovator in the age of agentic AI ChatGPT now also tries to respond based on old chat conversations, but not all of my conversations are about me. I'm often researching for articles and projects. These days, every research question may be interpreted as about me and attached to my personal dossier, when those questions might have nothing to do whatsoever with my life, other than a momentary bit of curiosity. And while the new interface shows its aggregated assumptions about who I am, I never really know what it thinks it knows about me, and therefore what assumptions it makes. As one of my high school chemistry teachers once said, "Assume means to make an ass out of you and me." While Dreaming V3 is clearly a technical triumph in breadth, efficiency, and scalability, I'm going to go so far as to say it's an irresponsible feature. First, it processes using old conversations, many of which occurred when the prevailing user perspective was that the AI only knew of the current conversation. Second, it's nearly impossible to prune what the AI recalls or decides about you. And third, V3 can't really keep up with your real life or changes, despite OpenAI's claims that it can. As much as AI is supposed to reduce cognitive load on humans, it actually increases our thinking burden when filtering out AI bias and hallucinations from responses. Now, we need to factor our entire history of conversations with the AI against every answer. But not all users will have the cognitive discipline to examine and verify the veracity and completeness of every AI response. Will AI conveniently leave out information because its warped view of who we are suggests we aren't interested in an area or approach? Will it modify its presentation because it thinks we only want to receive information in a certain format? We could never really trust an AI's answers. We knew that going in, back at the beginning of 2023. But now, with this memory capability, with this fundamentally selective memory capability, we can expect answers will probably be further skewed to match some aggregate internal representation of who it thinks we are, what it thinks we care about, and how it thinks we want to be communicated with. Also: I had ChatGPT build me a free PDF editor because I didn't trust it to change my files Then, of course, there are the monumental privacy considerations. We should always assume that cloud providers (Google, I'm looking at you) store every little detail of information about us for later use. But this level of profiling seems even more invasive and troubling. Misty water-colored memories Look at this bizarre conversation. I asked, "What is the single most important thing I care about?" My answer, unquestionably, is my wife and my little dog. More than anything else in the world, I love those two. But the AI says, "Preserving your ability to keep doing meaningful, independent work." Where the hell did it get that answer? Yes, I regularly interact with ChatGPT about the health of different foods and aspects of exercise, because I care about my health. I also talk to it about sustainability questions in light of things like Google's use of AI answers in place of presenting search results. But the most important thing? Far from it. The AI says it came up with that assessment because of the following conversational factors: * Web publishing resilience in the face of AI * Owned audience and newsletter growth * Effectiveness as a technology journalist * Technical independence * Remote-work viability * ZATZ Labs and WordPress projects * YouTube and maker work * Home lab, automation, backups, cameras, lighting, and workflow reliability Note that the AI didn't disclaim that its assessment was limited to these conversational factors. It has no knowledge, for example, that I've been working on Mac, iPhone, iPad, and Apple Watch programming projects in Claude Code. But the mere fact that it left Denise and Pixel out of my "most important" list tells me everything I need to know about the quality of ChatGPT's assumptions. They are skewed, inaccurate, out of date, often highly irrelevant, generalized from very specific temporary contexts, and potentially dangerous if used to limit informational responses. And that's why this new Dreaming V3 feature freaks me out. It should raise the hairs on the backs of your necks, too. Do you prefer an AI assistant that remembers your past conversations or one that treats each chat as a clean slate? Let us know in the comments below. You can follow my day-to-day project updates on social media. Be sure to subscribe to my weekly update newsletter, and follow me on Twitter/X at @DavidGewirtz, on Facebook at Facebook.com/DavidGewirtz, on Instagram at Instagram.com/DavidGewirtz, on Bluesky at @DavidGewirtz.com, and on YouTube at YouTube.com/DavidGewirtzTV.

ChatGPT now quietly rewrites its memories of you, and that's both good and unsettling

For decades, we've treated being known as a violation. We've gone on and on about de-Googling our lives because we couldn't stand the idea of a company knowing where we just went, what we bought, and who we texted. We installed the blockers, cleared cookies, and enabled every privacy toggle we could find. And then a chatbot asked us how our day was, and we decided to tell it everything. Funnily enough, the very reason that pushed people toward these tools is the same thing we would have flagged as creepy everywhere else: the mere fact that it remembers us -- our quirks, our projects, the offhand thing we mentioned three weeks ago and forgot. Every major AI tool does some version of this now, and while almost all of them call it memory, the way each tool's memory works under the hood is where they split. ChatGPT just updated how its memory works recently, and I'm not entirely sure I like it. ChatGPT now rewrites what it knows about you in the background A short history of ChatGPT learning you by heart OpenAI launched Memory in April 2024, which it also referred to as saved memories, which did exactly what it sounded like. You told ChatGPT to remember something, and it jotted it down. The catch was that anything you didn't explicitly tell it to keep, it forgot. It only saved things mid-chat, and only when you gave it a clear enough cue like "remember I'm traveling to Singapore in July." However, even that came with a second problem: the notes it kept didn't age. Tell it you're traveling to Singapore in July, and months later it still things you're going. In April 2025, the company improved ChatGPT's memory by enabling the model to reference chat context outside saved memories, which it called "dreaming." This was a method for ChatGPT to curate memories automatically in the background by pulling from your chat history. However, this system wasn't built to stand on its own and only caught context you'd never thought to flag. OpenAI itself said that it was never enough to carry memory by itself, and this is what the update in June 2026 changed. Want to stay in the loop with the latest in AI? The XDA AI Insider newsletter drops weekly with deep dives, tool recommendations, and hands-on coverage you won't find anywhere else on the site. Subscribe by modifying your newsletter preferences. The June 2026 update rebuilt how ChatGPT handles memory from the ground up. Instead of storing a static list of facts about you, ChatGPT now leans on dreaming as the whole foundation. It reviews your past conversations in the background, extracts what matters, tracks how your life has actually changed, and builds a running picture of who you are right now. The new system focuses on three things Recall, preferences, and not being stuck in the past With the June update, ChatGPT's new improved memory system focuses on three things. The first is carrying context forward. This basically means you tell it something once, and it holds onto it across every chat after. OpenAI's own example is shopping for camera gear. You mention your specific setup once, then ask months later what's compatible with "my photography setup," and instead of handing you a generic compatibility checklist to work through yourself, it already knows the housing and strobes you own and narrows it down to the actual parts that fit. OpenAI claims that factual recall has jumped from 41.5% with the old 2024 saved-memories setup to 82.8% under the new system. The second is following your preferences and constraints. If you've told it you're vegetarian, it stops suggesting meal preps with chicken in them. If you mention you live near San Francisco, it quietly assumes local options should be tailored to the area without you having to spell that out every time. OpenAI splits these into a few types: explicit instructions ("don't bring up Stan again"), hard constraints (the vegetarian one), and implicit preferences it infers on its own. The third, and what I think is most important, is staying current over time. Given that memory is supposed to account for time actually passing, this is the fix for the staleness problem I mentioned above. If you told ChatGPT you're planning a trip to Malaysia in August 2026, and then August comes and goes, the old system would still be talking about it like it's coming up. The new one revises it on its own and updates it to "went to Malaysia in August 2026" behind the scenes once the dates pass, so it stops planning a trip you've already taken and starts treating it as something behind you. While a trip is a trivial example, imagine if ChatGPT kept congratulating you on a job you'd just been let go from, or asking how things are going with someone you'd recently broken up with. The things memory gets stale about aren't always neutral, and a system confidently working off an outdated picture of your life can land somewhere between unhelpful and painful. So, this is an update that I think all AI tools could benefit from, not just ChatGPT. It's better than before, but memory worries me Better at remembering, worse at forgetting Now, the new system is an upgrade, and I felt it immediately. ChatGPT doesn't trip over things I corrected as much, doesn't recommend tools I'd already moved off, and I don't find myself re-explaining context over and over again. Every time ChatGPT responds, you can now also see exactly which past chats, saved memories, custom instructions, or files it leaned on to personalize that specific answer. You can even mark a source as relevant or not, or tell it to stop using that detail. These are all meaningful upgrades, and they all contribute to a much better experience. In the Settings app, you can view a Memory summary, which shows when it was updated and lets you edit what's there. You can do so by typing a correction into a box at the bottom, highlighting any specific line to fix it, or hitting Don't mention this again to stop ChatGPT bringing that detail up. Deals Save on AI software, subscriptions, and related deals now Explore discounts and limited-time offers on AI software, productivity apps, and subscription bundles. Find savings on privacy tools, cloud services, writing assistants, and creator suites -- plus accessories and service add-ons to streamline workflows and cut costs. Deals Explore Software, AI & Subscriptions Deals Beyond the fact that AI tools being able to add anything and everything into memory is unsettling on its own, the part that actually gets me is how little of it I can truly see or control. OpenAI's own help docs admit the memory summary won't include everything ChatGPT remembers, and that the system itself decides which details aren't appropriate to show in this view. So, the page that's built to let me audit and edit what it knows is a partial one. Speaking of which, there's also no single place to fully delete something, either. The Don't mention this again button only suppresses a detail, and to actually remove it from memory, you need to hunt it down across every chat, file, and connected app where it might live. Another worry I have is that users no longer get to decide what's worth remembering in the first place. The old system mostly waited for a cue. You'd explicitly tell it to "remember xyz thing" and it did. The new one is built to pick things up on its own without being asked. In some ways, this means the judgment call about what's worth keeping has moved from the user to the model. For the way most people use ChatGPT and AI tools, that's a lot. While you might not want it permanently remembering personal context, most people hand it over anyway. While you shouldn't be giving it confidential company information, most people paste in entire internal emails to come up with the best response. Both of these examples aren't stuff you'd want saved to a running profile of you. With this update though, that choice isn't really yours anymore. You typed it, and as far as the system's concerned, it counts. Another issue, which isn't really applicable to everyone, is what happens when more than one person uses the same account. I share my ChatGPT Plus account with my younger sister (since that account is only for testing purposes), and when I opened the memory summary, I wasn't looking at a picture of me. Instead, it was a confident blend of two different people folded into a single profile. It included details about my work tangled up with a completely separate set of interests and projects that weren't mine at all. At no point did it consider that two different people might be typing into the same box. It just smoothed us into one person! Unfortunately, whether we like it or not, it seems like every AI tool is heading in this direction. When you look at it from their perspective, it does make sense. Users seem to stick with tools that already know them. The more an assistant remembers, the more it costs you to leave. So, a memory that's hard to walk away from is how AI labs will clearly keep their users locked in.

ChatGPT's new memory builds a profile of you on its own -- and OpenAI admits you can't see all of it

A major memory upgrade could change the way the chatbot interacts with you every day OpenAI recently began rolling out a new memory system called Dreaming, designed to help ChatGPT build a more useful understanding of who you are over time. Instead of relying primarily on facts you've explicitly asked it to remember, the new system synthesizes information from your past conversations and updates those memories automatically as your life changes. The good news for users is that this feature is rolling out in stages. Dreaming reached ChatGPT Plus and Pro subscribers in the U.S. first, with free, Go and international users expected to follow over the coming weeks. So, depending on your plan and region, you may or may not see it yet. As the upgrade rolls out, ChatGPT may be remembering more about you than you realize. As soon as I learned about the update, I checked one setting immediately. The setting to review right now I recommend doing this right now: open ChatGPT and navigate to Settings > Personalization > Memory > Reference chat history If this setting is enabled, ChatGPT can draw on your previous conversations to personalize future responses. For example, if you've mentioned your favorite sports team, dietary preferences, travel plans, writing style or career goals across multiple chats, ChatGPT may use that information later without you needing to repeat it. The goal is to make conversations feel more natural and cut down on the context you have to provide. For many people, that's a feature. For others, it's something worth reviewing. What 'Dreaming' actually does According to OpenAI, Dreaming is a more advanced memory architecture that continuously synthesizes information from your conversations and keeps it current as circumstances change. The company's own example: a memory that reads "going to Singapore in July" rewrites itself to "went to Singapore in July 2026" once the trip is over, so the user then no longer needs to input that information. Seems fine, but here's the part that's easy to miss. Previously, ChatGPT's memory worked like a list of notes, in that it mostly remembered what you explicitly told it to. Dreaming shifts that work into the background, so the system now decides on its own what's worth keeping, including things it inferred about you and then quietly revised over time, not just facts you stated outright. That's a meaningful difference, and it makes the record harder to audit: OpenAI acknowledges that the memory summary page may not capture everything ChatGPT remembers about you. Memory is becoming somewhat of an evolving profile. As AI assistants mature, we're moving away from one-off conversations and toward tools that hold context across weeks, months and eventually years. OpenAI says Dreaming was built specifically to make ChatGPT more helpful by forming a more complete understanding of each user. That can be genuinely convenient. But it also means many users will want to spend a few minutes seeing exactly what information ChatGPT is drawing on to personalize its responses. On the bright side, OpenAI still gives you control. You can disable memory, manage or delete stored memories individually, and use Temporary Chat sessions that don't contribute to memory at all. My advice I'm not turning Dreaming off, and frankly, I don't think most people should. The whole point is that ChatGPT stops making you repeat yourself, and on that count it delivers. But convenience and visibility are pulling against each other here. The old memory was a list you were the one managing every time you wrote. The new one is a profile ChatGPT writes about you, partly from things it inferred rather than things you said, and OpenAI itself acknowledges the summary page may not show all of it. It's a good idea every so often to open your memory summary and actually read it. Don't just check whether memory is on, dive deeper and check whether it's right. A wrong fact you stated, you can spot and delete in seconds. A wrong assumption the system quietly made about you is the kind of thing that shapes every answer you get next, and the hardest to see. Follow Tom's Guide on Google News and add us as a preferred source to get our up-to-date news, analysis, and reviews in your feeds. Subscribe to Tom's Guide on YouTube and follow us on TikTok. Finally, you can visit our dedicated Tom's Guide Savings Squad hub for expert help on getting the best products for less.