Ask ChatGPT about running shoes on a Free account. Meanwhile, a shoe brand uploads a spreadsheet of customer emails and phone numbers to OpenAI. If you are on that sheet, you may be eligible for their ad only because of it. You never see the list, and you never get a say. Here is how the system works, what OpenAI says it protects, and a five minute settings check that puts you back in control of what you can control.
Here’s the thing. Most of us treat a chatbot like a private notebook. We paste in drafts, ask awkward questions and think out loud. That habit meets an advertising business the moment you use the Free or Go plan, and a new report shows the ad targeting goes further than most people assume.
Let’s break it down.
What 404 Media Found
On October 15, 2026, journalist Joseph Cox published a test he ran by buying ads inside ChatGPT. His finding: advertisers can aim at specific groups of people using an email address, a phone number or a unique identifier on a phone, on top of targeting based on what users type into the chatbot. You can read the opening of the story on 404 Media, and the full piece sits behind a free account.
Gizmodo’s write-up fills in the mechanics. An advertiser uploads a list of people. OpenAI checks which entries link to ChatGPT accounts. Matches become eligible to see the ad.
404 Media itself notes this is fairly standard across the ad industry. What makes it feel different is the setting. Sensor Tower estimated that the ChatGPT app passed 1 billion monthly users in May, according to Reuters. Plenty of those people type things they would never type into a search bar. The same outlet also reported that human contractors read some real conversations in an effort called Project Lily. Prompts are less private than they feel.
How ChatGPT Picks Your Ad
OpenAI’s Ads in ChatGPT help page lists the inputs. Here is the short version.
| Signal | When it is used |
|---|---|
| Your current chat thread | Always, even with personalization off |
| The ad’s landing page, plus the advertiser’s context hints and targeting choices | Always |
| General location and language | Always |
| Past chats, memory, and how you interact with ads | Only if ad personalization is on |
| An advertiser’s uploaded list (a custom audience) | When the advertiser chooses it. OpenAI’s pages do not say how the personalization toggle affects it |
Context hints are the part worth understanding. Advertisers describe topics or situations where their product fits, and OpenAI decides the final match based on your conversation. So the prompt you type is a targeting signal. It stays inside OpenAI’s systems, according to the company, but it shapes what shows up under your answer.
The Email and Phone Part, Explained
OpenAI’s custom audiences guide for advertisers spells out the rules. This is what an advertiser can do and what limits it.
- Upload a file of emails, phone numbers, SHA-256 hashed versions of both, or Google Advertising IDs, which are the ad identifiers on Android phones.
- Use that list to include people in a campaign, exclude them from one, or raise and lower bids for them by a factor between 0.1x and 10x.
- Include only if at least 25,000 uploaded identifiers match real ChatGPT users. OpenAI recommends 100,000 or more. Smaller lists can be used only to exclude people.
- See no individual matches. The Ads Manager does not show who matched and does not let an advertiser pick specific people.
- Rely on OpenAI deleting the uploaded file after processing, usually within 24 hours.
What this really means is that a brand cannot single you out. It can target a crowd you belong to, and that crowd may be built from your personal data. Both things are true at once.
The entry price is also low. OpenAI’s campaign setup page lists a minimum daily budget of 25 USD, 15 EUR or 15 GBP. Older guides that mention enterprise only pricing are out of date.
What OpenAI says stays off limits
- Advertisers never receive your chats, chat history, memories, name, email, precise location or IP address. They get totals such as views and clicks.
- OpenAI says it does not sell your data and that ads do not influence ChatGPT’s answers.
- Ads do not appear in Temporary Chats, for accounts identified as under 18, or near sensitive topics such as personal health, mental health and politics. Political advertising is not allowed.
- Plus, Pro, Business, Enterprise and Edu accounts do not see ads.
Where the Gaps Are
I read OpenAI’s pages closely, and three questions remain open.
Does the personalization switch cover advertiser lists? OpenAI says turning personalization off stops it from using other chats, ad history and topics. Gizmodo points out the setting’s wording covers your past chats, activity and preferences, which does not clearly say you drop out of an advertiser’s audience. I could not find OpenAI answering this directly.
Is raw data hashed on upload? Per Gizmodo’s reading of 404 Media’s report, it is unclear whether OpenAI hashes the plain emails and phone numbers advertisers upload. OpenAI’s answer is that files are deleted within about a day.
Does turning personalization off remove ads? No. OpenAI states that you will still see ads based on your current chat.
Step by Step: Lock Down Your ChatGPT Ad Settings
This takes about five minutes. The steps follow OpenAI’s help page, and menu names can shift, so update the app if something is missing.
- Check your plan. Only Free and Go accounts see ads. If you pay for Plus or Pro, you can stop here.
- Open Settings, then Ad Controls. The section appears only in regions where ads are live or coming soon. If you do not see it on mobile, update the app first.
- Turn off Personalize ads. This stops ChatGPT from using other chats, ad history and topics to choose ads.
- Turn off Past chats and memory in the same screen. With it off, ads lean on your current chat only.
- Review your history and topics, then use Delete ads data. This does not touch your chats. OpenAI says full removal can take up to 30 days.
- Use the three dot menu on any ad. Choose About this ad to see why it showed up, Hide ad to dismiss it, or Report ad if it looks misleading.
- Switch to Temporary Chat for anything personal. No ads appear there, and it does not use or update memory.
Three Extra Moves
- Go ad-free on Free. Open Settings, then Ads controls, tap Change plan to go ad-free, and pick Reduce message limits. The trade is real: fewer daily messages and no image generation or deep research. Plus and Pro remove ads without those cuts.
- Delete your Android advertising ID. On Android, go to Settings, tap Privacy, then Ads, and choose Delete advertising ID. Google’s help page warns that apps can still use other identifiers. This only affects matches built on that ID, not matches built on your email or phone.
- Think about which email and phone you use to sign up. Advertiser lists match on what advertisers hold. If your ChatGPT login uses a different address from the one you give retailers, a match is less likely. That is my reasoning, not OpenAI guidance, so treat it as a bet and not a guarantee.
A Rule of Thumb That Actually Helps
Treat the free tier like a search engine, not a diary. If you would not type it into Google next to a shopping tab, use a Temporary Chat or skip the cloud tool. That rule costs nothing and works whatever the fine print says.
If ads are a dealbreaker for you, another option exists. Anthropic said in February that Claude will remain ad-free. To keep this fair, 404 Media’s reporting also notes that Anthropic confirmed it uses human review to improve its models. No ads does not mean no data questions. Check the data controls on any chatbot you use.
Want a wider cleanup? My guide on finding your exposed phone and email online and filing removal requests pairs well with this one.
Honest Limitations
- Most privacy claims come from OpenAI. Deletion timelines, no data sales and no sharing with advertisers are the company’s statements, and I found no independent audit.
- Ads are still a test that keeps growing. OpenAI began testing in the US on February 9, 2026, and availability varies by region. Personalized ads are not initially available in the EEA or Switzerland.
- Settings can change. OpenAI’s help page showed an update only minutes before I read it. Confirm each step on your own screen.
- The email and phone match is documented, not fully explained. The open questions above remain until OpenAI answers them.
- I could not read all of 404 Media’s story. Details beyond its opening paragraphs come from OpenAI’s documentation and from Gizmodo’s coverage of the report.
The Bottom Line
You cannot be singled out by a brand, and your chats are not handed to advertisers. That is what OpenAI says, and its own documentation backs the audience limits. But your email, phone number or phone ID can still decide which crowd you stand in, and your current prompt always counts as a signal.
Take the five minutes. Open Ad Controls tonight, switch off personalization, clear your ads data, and move anything sensitive to Temporary Chat.
Know someone who uses ChatGPT for everything? Send them this post before their next prompt.
Sources
- 404 Media: How ChatGPT Targets Ads To You Based On Your Prompts, Phone Number, and Email
- Gizmodo: Here’s How ChatGPT Is Targeting You With Ads
- OpenAI Help Center: Ads in ChatGPT
- OpenAI Help Center: Set up Custom Audiences for your Campaign
- OpenAI Help Center: Create Campaigns for ChatGPT Ads
- Reuters: ChatGPT app hits 1 billion monthly active users in record time
- 404 Media: Inside Project Lily, the Humans Reading Your ChatGPT Chats
- Google Play Console Help: Advertising ID
- The Decoder: Anthropic pledges to keep Claude ad-free
- ForkLog: 404 Media on contractors reading ChatGPT chats
Leave a Reply