Does ChatGPT Use Your Past Chats For Ads? OpenAI's $1 Billion Expansion, Explained
ChatGPT can use the current conversation—and, with personalization, past chats and memory—to select ads. Here is what advertisers receive and how to opt out.
Yes—ChatGPT can use your current conversation to choose an ad, and OpenAI says past chats and memory may also be used when Personalized Ads is enabled and available in your country. That does not mean advertisers receive your conversations. OpenAI says the matching happens inside ChatGPT and advertisers do not get your chats, chat history, memories, personal details, precise location, IP address, or sensitive information.
Watch The 30-Second Summary
Research cutoff: August 31, 2026, 11:30 PM America/Chicago. Ad availability, controls, and plan eligibility can vary by country and can change as OpenAI expands the program.
OpenAI announced on August 31 that ChatGPT Ads reached a $1 billion annualized revenue run rate in less than 200 days. The company says tens of thousands of advertisers now use the platform, ChatGPT Ads are available through its ecosystem in more than 40 countries, and self-service Ads Manager access is expanding across India, Europe, the Middle East, and North Africa.
The scale makes a once-theoretical privacy question immediate: when an ad appears next to a private AI conversation, which parts of that conversation are used, who sees them, and which controls actually change the data path?
The short answer has two parts:
- OpenAI's ads system can use context inside ChatGPT to select an ad.
- OpenAI says the advertiser does not receive the underlying chat or personal profile.
Those statements can both be true. “Private from advertisers” is not the same promise as “not used by the platform's ad-selection system.”
What Is Confirmed
The Current Chat Can Affect Which Ad You See
OpenAI's announcement says its ads system uses the context of the current conversation to show an ad relevant to what a person is exploring. Its current Ads in ChatGPT help page says ad matching may use what you are discussing in the current thread, plus basic context such as general location or language.
This is contextual advertising. If you ask about renovating a kitchen, learning Spanish, comparing laptops, or planning a trip, the topic of that live conversation can affect the ad shown below the answer.
Turning off Personalized Ads does not necessarily make the current conversation irrelevant to ad selection. OpenAI says that control makes ads less tailored; it does not remove ads or disable every contextual signal.
Past Chats And Memory Can Be Used When Personalization Is On
OpenAI's help page says Personalized Ads can add several signals:
- the current chat thread, including personalized model responses;
- ads you hide or engage with;
- past chats; and
- memory.
OpenAI's August 31 announcement phrases this more broadly: depending on country and settings, the ads system may use context from a person's wider ChatGPT experience.
That means an ad beside today's answer may reflect more than today's prompt when personalization is available and enabled. The system may draw on broader interests or preferences inferred from prior ChatGPT activity.
Advertisers Do Not Receive Your Chats, According To OpenAI
OpenAI draws a sharp line between information used internally for ad selection and information delivered to an advertiser. Its help documentation says advertisers never receive:
- chats or chat history;
- memories;
- name or email address;
- precise location or IP address; or
- sensitive information such as health, mental-health, or political topics.
The company says advertisers receive aggregate reporting such as views and clicks. Advertisers can provide context hints describing conversations, topics, or keywords where their products may be useful, but OpenAI says those hints are not exact-match keywords and do not give the advertiser access to the conversation.
So, seeing an uncannily relevant ad is not proof that the advertiser read your prompt. It can result from OpenAI matching an ad to conversation context without disclosing the text to the brand.
Ads Are Supposed To Stay Separate From Answers
OpenAI says advertising does not influence ChatGPT's answers. Ads are labeled and visually separated from the organic answer.
The current ad policies also prohibit ad placements near categories OpenAI considers inappropriate or sensitive. The policy names privacy, political content, mental and personal health conversations, emotionally reliant interactions, and other sensitive user journeys.
Those are stated product and policy boundaries. They are important, but they are not an independent audit proving that every live classification and placement is always correct.
Users Have More Than One Ads Control
OpenAI documents four different actions that are easy to confuse:
| Control | What it changes | What it does not automatically do |
|---|---|---|
| Turn off Personalized Ads | Stops broader personalization signals such as past chats and memory from tailoring ads | Does not necessarily remove ads or stop contextual matching to the current chat |
| Delete ads data | Stops deleted ads-history and topics data from being used for ads | Does not delete chats; server removal may take up to 30 days |
| Use a Free no-ads option where offered | Removes ads in exchange for lower message limits and reduced tool access | Does not create a general privacy mode for every other ChatGPT data setting |
| Upgrade to an ad-free plan | Removes ads on eligible Plus or Pro plans | Does not by itself turn off memory, chat history, or model-improvement settings |
OpenAI says controls may vary by region and eligibility. If your account participates in the program, look under Settings → Ads controls and read the wording shown for your plan and country.
What Is Still Unclear
How Broadly Past-Chat Personalization Is Enabled
OpenAI says the answer depends on country and user settings. The August 31 announcement covers a global advertiser expansion, but it does not provide one universal matrix showing exactly which consumer accounts currently use past chats, memory, only current-chat context, or no ads at all.
Do not infer your configuration from another user's screenshot. Check the controls on the exact account, plan, app, and country you use.
How The System Converts A Conversation Into An Ad Signal
OpenAI lists the categories of signals that can be used, but it does not publish the full ranking model, weighting, inference process, sensitivity classifier, or every topic label that can be assigned.
There is a meaningful difference between sending a raw prompt to an advertiser and generating an internal topic or relevance score. OpenAI says it does the latter and not the former. Users still cannot independently inspect every feature the system derived from a chat or reconstruct why one eligible advertiser won the auction.
Whether Every Sensitive Conversation Is Classified Correctly
OpenAI's policy says ads should not appear near privacy, mental-health, personal-health, political, and other sensitive contexts. Real conversations are often ambiguous. A discussion about running shoes can become a medical conversation; a budgeting question can reveal financial distress; travel planning can expose a religious observance, disability, or family emergency.
The public policy identifies the intended boundary. The August 31 announcement does not provide error rates for sensitive-context detection across languages, dialects, indirect disclosures, long chats, or mixed-topic threads.
How The Expanded Platform Performs Across Different Users
A new primary study, The Beginning of ChatGPT Ads, audited more than 3,000 ads shown to 91 synthetic U.S. accounts. The researchers reported that lower-income synthetic accounts were more likely to receive ads and that the early ads were clearly separated from answers and skewed toward consumer goods.
That is useful independent evidence about an early U.S. deployment. It is not proof about every real user, every demographic, the newly expanded international markets, or the production system after later changes. The study used controlled sock-puppet accounts and should be read as an audit of a specific period and design, not a universal discrimination finding about every placement.
Five Privacy Questions To Ask About A ChatGPT Ad
Use the AD PATH check when an ad feels unusually personal.
A — Account And Plan
Confirm which account is signed in, the plan, the country, and whether it is a personal or managed workspace. Ad eligibility differs across plans and regions.
If you use more than one browser profile or account, do not assume they share the same settings.
D — Data Signals
Open Ads controls and inspect whether Personalized Ads is on. Review available ad history and topics. If an ad has a “Why this ad?” or similar explanation, read it before guessing which chat caused the placement.
Remember that the live conversation can remain a contextual signal even when broader personalization is off.
P — Personalization, Memory, And Training Are Separate
Do not treat one toggle as a master privacy switch.
- Ads personalization controls which broader ChatGPT signals may tailor ads.
- Memory controls whether ChatGPT saves or references information to personalize answers.
- Reference chat history controls whether past conversations can inform future responses.
- “Improve the model for everyone” controls whether eligible conversations contribute to model improvement.
- Chat deletion and retention govern how long conversation records remain.
Changing one does not silently guarantee the others changed.
A — Advertiser Access
Separate relevance from disclosure. OpenAI says advertisers get aggregate performance information, not your prompt, history, memory, identity, precise location, IP address, or sensitive information.
An ad that matches the topic is evidence that some matching occurred. It is not, by itself, evidence that the brand received the chat.
If an ad links to a site, clicking creates a new data path. The destination can receive the ordinary information associated with visiting a website, and conversion tracking may measure what happens after the click. Read the advertiser's privacy policy before entering personal data.
T — Turn Off Or Delete The Right Thing
If you do not want past chats or memory used to tailor ads, turn off Personalized Ads where the control is available. If you want the ads-history and topics profile cleared, use the separate delete-ads-data control. OpenAI says full server removal may take up to 30 days.
If you want no ads, choose an eligible no-ads plan or reduced-limit Free option where offered. If you want chats deleted, use chat or account deletion controls separately.
H — Handle Sensitive Prompts Before Sending
Ad controls do not make a hosted AI service appropriate for every secret. Remove unnecessary names, credentials, account numbers, health identifiers, client data, and other details you are not authorized to send.
For medical, legal, financial, employment, school, or regulated work, follow the rules that apply to the information. A promise that advertisers cannot read chats is not a promise of legal privilege, regulatory compliance, zero retention, or no provider processing.
What This Expansion Does Not Prove
OpenAI's $1 billion milestone does not prove that:
- advertisers can read private ChatGPT conversations;
- ads change the text of ChatGPT's answers;
- turning off ad personalization removes all ads;
- disabling model training disables ad matching;
- paying for an ad-free plan deletes old chats or memories;
- every country uses past chats for ads in the same way;
- every sensitive conversation will always be classified correctly;
- clearing ads data immediately removes every server copy;
- clicking an ad keeps the destination from collecting normal website data; or
- an OpenAI policy statement is the same as an independent technical audit.
The defensible conclusion is narrower: ChatGPT's own ad system can use conversation context, and it can use past chats and memory when the relevant personalization setting and country rules allow it. OpenAI says those inputs stay private from advertisers.
Where OpenVeil Fits—And Where It Does Not
OpenVeil is a hosted, privacy-focused AI workspace for adults. Normal chat history stays in the user's browser rather than being kept as a normal server-side chat-history record. OpenVeil does not use prompts, uploads, images, audio, selected local history context, or AI outputs to train foundation models.
That offers a different history boundary for people who do not want a normal cloud account archive of their AI conversations. It does not mean OpenVeil is fully offline or anonymous. Active prompts, searches, files, voice, image, video, and other requested features still require transient processing by OpenVeil and necessary AI, search, upload-processing, hosting, routing, security, billing, and infrastructure providers.
OpenVeil does not change ChatGPT settings, block ChatGPT ads, protect an advertiser's website, or guarantee that sensitive information is safe to send to any hosted service. Browser-local history can also be lost by clearing browser data or changing devices and may still be accessible to someone who controls the device or browser profile.
If that narrower history model fits your work, you can try OpenVeil with a one-time 10-action preview that does not require a card. Before sending sensitive material, read What Browser-Local Chat History Means In An AI App and What To Check Before Trusting Any AI Privacy Claim.
Frequently Asked Questions
Does ChatGPT Use Past Chats To Show Ads?
It can when Personalized Ads is enabled and that form of personalization is available for your country and account. OpenAI says the current chat can be used for contextual matching even without broader personalization.
Can Advertisers Read My ChatGPT Conversations?
OpenAI says no. Its current documentation says advertisers do not receive chats, chat history, memories, personal details, precise location, IP address, or sensitive information. They receive aggregate ad-performance reporting such as views and clicks.
Does Turning Off Personalized Ads Stop All Ads?
No. OpenAI says turning personalization off can make ads less tailored, but it does not remove them. Eligible users may have a separate reduced-limit Free no-ads option, and OpenAI says Plus and Pro are ad-free.
Does Deleting Ads Data Delete My Chats?
No. OpenAI says deleting ads data clears ads history and topics data without affecting chats. It may take up to 30 days for that ads data to be fully removed from OpenAI's systems.
Can Ads Appear Next To Health Or Political Conversations?
OpenAI's current policy says ads should not appear near personal-health, mental-health, political, privacy, emotionally reliant, and other sensitive contexts. That is the stated policy; the company has not published a universal error rate proving every conversation is always classified correctly.
Are ChatGPT Answers Influenced By Advertisers?
OpenAI says ads do not influence answers and are clearly labeled and separated from the response. The advertiser can supply ad copy and context hints, but OpenAI says those hints do not determine the organic answer.
Is An Ad-Free ChatGPT Plan The Same As A Private Chat?
No. Removing ads is one control. Memory, reference chat history, model improvement, retention, deletion, connected apps, files, legal exceptions, and workspace administration are separate data paths.
Is OpenVeil Fully Offline Or Anonymous?
No. OpenVeil is hosted. Its relevant documented distinction is browser-local normal chat history and no normal server-side chat-history record. Active requests still require processing by OpenVeil and necessary providers, and OpenVeil maintains the account, billing, security, and operational data needed to run the service.
Bottom Line
ChatGPT Ads now has the scale of a major advertising platform, and the privacy boundary is more nuanced than “the advertiser sees your chat” or “your conversation is never used.”
OpenAI says its system uses the current conversation to match ads and may use past chats and memory when Personalized Ads is enabled. It separately says advertisers do not receive those conversations, memories, personal details, precise location, IP address, or sensitive information.
If you use an eligible ad-supported account, check the exact Ads controls on that account. Turn off personalization if you do not want broader ChatGPT history tailoring ads, clear ads data if you want that separate profile removed, and choose a no-ads option if you do not want ads at all. Then review memory, model-improvement, chat-retention, and deletion settings separately.
Most importantly, make the privacy decision before the prompt is sent. No ad toggle changes the fact that a hosted assistant must process the information you submit.
Sources
- OpenAI: A Milestone In Expanding Access To AI
- OpenAI Help Center: Ads In ChatGPT
- OpenAI: Testing Ads In ChatGPT
- OpenAI: Ad Policies
- OpenAI: ChatGPT Privacy Settings
- OpenAI Help Center: Data Controls FAQ
- OpenAI Help Center: Chat And File Retention Policies In ChatGPT
- The Beginning Of ChatGPT Ads (Primary Research)
- Axios: OpenAI Says Ads Business Reached A $1 Billion Run Rate