Yes, as of August 2026 ChatGPT can use your saved memory to help decide which ads you see. When ad personalization is on, OpenAI can draw on the topic of your current conversation, your past chats, and the memories ChatGPT has stored about you to pick more relevant ads. This is the answer to the question a lot of people are now typing: does ChatGPT use my chats for ads, and can I turn it off. You can. Below is what changed, what actually feeds the targeting, why targeting is not the same as training on your data, and the exact settings to limit or switch it off.
The short version: an assistant you told personal things to now has a business reason to remember them. That does not mean your chats are being sold or handed to advertisers. It means the signals you gave ChatGPT, including memory you set up to make it more helpful, can now shape the ads inside the same app. Knowing the difference is the whole point of this piece.
What actually changed in August 2026
OpenAI expanded ads in ChatGPT to more countries and updated its privacy policy so that ad personalization can reference your memory and chat history. On August 11, 2026, the ad program rolled out to the United Kingdom, Mexico, Brazil, Japan, and South Korea, on top of the earlier United States testing. Alongside that expansion, the privacy policy was revised so ad selection has a documented data flow rather than an informal test.
The timeline matters, because ads did not appear overnight. Testing in the United States began in February 2026 on the free and lower tiers. The United States privacy policy picked up ad related data handling around April 30, 2026. A separate update on June 2, 2026 addressed European Union and United Kingdom requirements. The August 11 step is when the map got noticeably bigger and the memory link became explicit in the policy.
One honest caveat. Ads in ChatGPT are still a rollout in progress, not a finished, universal feature. Availability, the exact tiers that see ads, and the default settings differ by country. So treat the specifics below as what the current policy allows, not a promise about what every user everywhere will experience on any given day.
Targeting language, not a leak: OpenAI's position is that advertisers do not get access to your chats, chat history, memories, or personal details. The signals are used inside OpenAI's systems to choose ads. That is a real and important distinction from selling your data.
What signals feed the ad targeting
Three things can feed ad personalization: the topic of your current conversation, your past chats, and your saved memory. Your history of interacting with ads can be added to that mix. In plain terms, if you have told ChatGPT that you are planning a move, training for a marathon, or shopping for a specific product, those facts can influence which sponsored results you see later.
- Current conversation: the subject you are discussing right now can be matched to relevant ads in that session.
- Past chats: earlier conversations across sessions can be referenced when ad personalization is enabled.
- Saved memory: the personal details ChatGPT has stored about you, for eligible users, can be used as a signal.
- Ad interaction history: which ads you dismissed, clicked, or engaged with can shape what comes next.
The memory piece is the part worth slowing down on. Memory exists so the assistant feels less repetitive and more useful. You set it up to be remembered. The change is that the same store of personal facts now sits next to an ad system. Nothing was necessarily taken from you, but the purpose of that memory quietly widened.
Targeting is not the same as training your data
This is the confusion to clear up: using your memory to target ads is a different thing from training the model on your data. Ad targeting means your signals help select which ad to show you inside the app, in the moment. Model training means your conversations are used to teach the underlying AI for future versions. They are separate policies with separate controls, and mixing them up leads people to the wrong conclusion.
Why the distinction matters in practice: you can be opted out of training and still have ad personalization on, or the reverse. Turning off one does not automatically turn off the other. So if your goal is to keep your memory from shaping ads, you need the ad personalization control specifically, not just the training toggle.
If privacy is your reason, check both settings. Look for the training or improve the model control and the ad personalization control as two separate switches, because being out of one does not opt you out of the other.
How to turn off ChatGPT ad personalization and review your memory
You can turn off ad personalization in Settings, and you can review or delete what ChatGPT remembers about you. Ad personalization can be toggled on or off at any time, and ad related data can be deleted. If you switch personalization off, OpenAI stops referencing your other chat threads and ad history when choosing ads, though basic context based ads tied to the current conversation may still appear.
- Open Settings and find the ads or personalization section, then turn off ad personalization so past chats and memory are no longer used to select ads.
- Delete your ad data from the same area if you want to clear the history that has already built up.
- Open the memory or personalization settings to see what ChatGPT has saved, then edit or delete individual memories you would rather it forget.
- Check your data controls for the separate training or improve the model setting if you also want your chats excluded from future model training.
- Note the regional default: in the United States ads are on by default and you opt out, while in the European Union and United Kingdom personalized ads generally require you to opt in first.
One more option worth knowing: on the free tier there is a full ad opt out that trades personalized content for fewer daily messages. That is the honest cost of a free product. Fewer ads usually means fewer free messages, because the ads are what pays for the free usage in the first place.
What advertisers can and cannot see
Advertisers do not get your chats, your chat history, your memories, or your personal details. OpenAI's stated model is that the targeting happens internally: the signals stay inside its systems and are used to match ads to you, without exposing the raw conversation. Advertisers receive aggregated, non identifying performance information such as total views or clicks.
There is also a structural point in OpenAI's favor here. There is no third party cookie and no identity graph being passed around an ad exchange. The personal signals do not leave to a network of brokers. That is genuinely better than the open web ad model, and it is fair to say so. The concern is narrower and specific: a system you confide in now has a commercial incentive to use what it remembers.
ChatGPT memory ads versus a private memory approach
The table below separates what happens to your memory under ChatGPT with ad personalization on from a memory tool built without an ad business, so you can see where the lines differ.
| Your memory | ChatGPT with ads on | A private memory approach like MemX |
|---|---|---|
| Used to target ads | Yes, current chat, past chats, and saved memory can feed ad selection | No, there is no ad business model |
| Shared with advertisers | No, signals stay inside OpenAI and advertisers get aggregated stats | No, memory is isolated per user |
| Used to train the model | Separate policy and separate toggle from ads | No training on your data |
| Who can read it | OpenAI systems, with the controls it provides | You, with per user isolation and customer managed keys |
| How to limit it | Turn off ad personalization and delete ad data in Settings | Private by architecture by default, nothing to opt out of |
The honest bigger picture
Free products get paid for somehow, and ads are an old, understandable answer. If a service is free to hundreds of millions of people, the money comes from ads, subscriptions, or both. None of that is a scandal by itself. The specific shift here is that memory and ads have been connected. An assistant designed to remember your life now uses that memory to sell against it, inside the same window where you asked it something personal.
That is the line being crossed, and it is worth naming without panic. It does not mean ChatGPT is spying or leaking. It means the incentives changed. The more you tell it, the more valuable your profile becomes to the ad system, even if no advertiser ever sees a word you wrote. For some people that trade is fine. For others, especially with health, money, or family topics, it is a reason to keep the truly personal memory somewhere that has no ad incentive at all.
There is also a settings fatigue problem to be honest about. Controls that live several menus deep, split across ad personalization here and training there, are technically opt outs but practically ignored by most people. Defaults win. When ads are on by default in a country, the majority of users keep them on simply because they never open the panel. That is why the default itself, not just the toggle, is the thing to watch as this expands to more markets over the coming months.
Where MemX fits
MemX is built so that an assistant remembering you does not turn that memory into ad inventory. It stores your documents, photos, voice notes, and messages so you can ask a question and get an instant answer with the source: snap it now, find it later. The difference relevant to this post is the business model. MemX has no ad business, so your memory is never a targeting signal.
It is private by architecture: per user isolation, customer managed keys, encryption at rest, and on device processing where possible, with no training on your data. That is not a claim of end to end encryption or zero knowledge, and it is not a medical compliance promise. It is a straightforward structural fact: there is no ad system on the other side of your memory, so the incentive that connects your saved details to advertising simply does not exist here.
A practical split: keep general questions in whatever assistant you like, and keep the genuinely personal memory, health notes, finances, family details, in a tool with no ad model. Then no toggle can quietly reconnect it to advertising later.
01Does ChatGPT use my chats and memory for ads?
When ad personalization is on, yes. As of August 2026, OpenAI can use the topic of your current conversation, your past chats, and your saved memory to select more relevant ads. You can turn this off in Settings, though basic context based ads tied to the current chat may still appear.
02How do I turn off ChatGPT ad personalization?
Go to Settings and find the ads or personalization section, then switch ad personalization off. You can also delete your ad data there. With it off, OpenAI stops using your other chat threads and ad history to choose ads, though the current conversation can still provide basic context.
03Are advertisers seeing my private conversations?
No. OpenAI states advertisers do not get access to your chats, chat history, memories, or personal details. Targeting happens inside OpenAI's systems, and advertisers receive only aggregated, non identifying stats like total views or clicks. The concern is the internal use of memory, not an external leak.
04Is ad targeting the same as training the model on my data?
No. Ad targeting decides which ad you see in the moment using your signals. Model training uses conversations to improve the AI itself. They are separate policies with separate controls, so opting out of one does not opt you out of the other. Check both if privacy is your goal.
05How is MemX different from ChatGPT here?
MemX has no ad business, so your memory is never used to target ads. It is private by architecture, with per user isolation, customer managed keys, encryption at rest, and no training on your data. There is no ad system on the other side of your memory, so nothing to opt out of.
