Every time you sign up for a membership card, fill out a real estate inquiry form, or create an account on an unfamiliar website, your information doesn’t just disappear—it gets harvested and sold by companies known as Data Brokers (simply put, companies that specialize in collecting individuals’ personal data and reselling it to third parties). Searching for and requesting removal from each site on your own is nearly impossible, as there are hundreds of such sites across the internet.
After reading this guide, you will know exactly: how to find the websites currently displaying your information, how to use ChatGPT to draft legally sound opt-out/removal request emails, how to track them until everything is resolved, and how to utilize agent mode (a specialized mode where ChatGPT can browse the web and fill out forms on your behalf) to cut down most of the manual effort. All prompts in this post come with a ready-to-use copy button, so there is no need to retype them.
1. Setting Expectations: What ChatGPT Can and Cannot Do

Before you begin, you need to understand one thing clearly: ChatGPT cannot “erase all traces of you from the entire internet.” No tool (including professional paid services) can guarantee wiping every website, database, and backup on the web. Understanding this limitation helps you set reasonable expectations and prevents you from getting discouraged along the way.
What ChatGPT truly excels at is acting as a process assistant: it helps you search faster, draft professional texts, and avoid missing any steps. Specifically, ChatGPT can:
- Find public pages that are likely displaying your personal information.
- Compile a list of websites, links, and the types of information appearing in each location.
- Locate contact pages, privacy policies, or data removal request forms.
- Draft removal request emails or messages with a professional tone and appropriate legal grounding.
- Help you build a tracking sheet: which websites have responded, and which have not.
- Set up recurring check reminders if your account has an automated task feature (called Tasks: simply understood as “setting a schedule for ChatGPT to automatically remind you of a task at a specific date and time”).
- Assist with direct browser actions if your account has the agent feature—simply understood as a mode that allows ChatGPT to open the web, click buttons, and fill out forms on your behalf, just as if you were controlling the computer, but done by ChatGPT instead. Section 6 of this document will detail how to enable and use this mode.
In short: ChatGPT is not an “internet eraser,” but it is an assistant that makes your data removal process faster, more organized, less prone to oversights, and, when needed, almost entirely automated using an agent.
2. When should you audit your personal data online?
You should set aside a session to check your personal information online if you fall into any of the following cases:
- You have received numerous spam calls, financial scam messages, or communications impersonating authorities.
- You have registered accounts across multiple e-commerce platforms, forums, or legacy apps that haven’t been used in a long time.
- You work in specialized fields (law, finance, journalism, systems administration) that require personal privacy protection.
- You are starting to look for a new job or preparing to launch your own business.
When searching, make sure to thoroughly audit the following four groups of information, ranging from less sensitive to highly sensitive:
- Group 1 (General public information): Full name, social media profile pictures, professional bio on LinkedIn.
- Group 2 (Secondary contact information): Secondary emails, forum usernames, old social media handles.
- Group 3 (Direct contact information): Primary email, personal phone number, workplace address.
- Group 4 (Highly sensitive data): National ID/Citizen ID numbers, home address, photos of identity documents, financial/health information, internal company data.
If time is limited, prioritize handling: phone numbers, home addresses, photos of identity documents, financial details, and exposed customer data. These carry the most immediate risk of harm if exploited by malicious actors.
3. Security Setup and Safe Keyword Preparation
Before asking ChatGPT to search, prepare a list of keywords about yourself to get more accurate results. For example, a well-prepared set of keywords includes:
- Full name with accents and without accents (e.g., “Nguyễn Văn A” and “Nguyen Van A”).
- Personal or work email addresses you have used.
- Usernames/nicknames commonly used on social networks and forums.
- Company or school names combined with your full name.
Never type into the chat box: passwords, OTP codes, national ID numbers, bank account numbers, or photos of personal ID documents. If you need to search for sensitive information, only describe the type of information instead of sending actual data.
Before entering any data, go to Settings → Data Controls on ChatGPT and turn off the “Improve the model for everyone” option. This action prevents OpenAI from using your information to train future public AI models.
4. Legal Basis: Decree 13/2023/ND-CP
If you simply message a website asking to “delete my information,” most will ignore it due to the lack of binding legal grounds. In Vietnam, collecting and displaying personal data without consent constitutes a violation of the law under Decree 13/2023/ND-CP on Personal Data Protection (effective since July 1, 2023) [1].
This serves as your essential “legal weapon” to include in every email or formal data removal request. A standard removal request must ensure three key elements:
- Accurate citation of legal documents: Decree 13/2023/ND-CP regarding the data subject’s right to request data deletion.
- A clear response deadline: Typically between 48 hours and 7 working days.
- Next steps if unaddressed: Establishing a bailiff’s deed (vi bằng) to record evidence or filing a complaint with the Authority of Information Security (Ministry of Information and Communications).
Some websites may ask you to provide identity documents to verify that you are the legitimate data subject before proceeding with the deletion. This is a lawful requirement under Decree 13 to prevent impersonation, but you should only send a scanned copy with non-essential details redacted, and solely via trusted portals.
5. 10-Step Process for Manually Removing Personal Information
Below is the 10-step process in logical order. Each step includes specific actions, accompanied by standard command templates and prompts so you can perform them without requiring complex technical knowledge.
Step 1: Perform an initial self-check on Google using exact match syntax
Open Google and search using the phrases below. Always enclose them in double quotation marks so Google searches for the exact phrase rather than individual words scattered across the page:
For example: typing “John Doe” (with quotation marks) filters out thousands of noisy, irrelevant results compared to searching without quotes.
Step 2: Have ChatGPT Search by Keyword Groups
Paste the prompt below into ChatGPT after filling in your information. Click the Copy button to get the verbatim text:
Step 3: Categorize Results and Assess Risk Levels
Once you have the list of links, categorize them into 4 levels: Urgent (financial information, citizen ID card, address), Important (phone number, primary email), Moderate (old forum usernames), and Shared-name information that requires verification.
Step 4: Create a tracking sheet for removal requests
Use the prompt below to have ChatGPT generate a detailed tracking sheet. You can copy this table to Google Sheets, Excel, or Notion to manage the status for each website:
Each website should have a clear status: Request not sent, Sent 1st time, Responded, Successfully removed, Removal rejected, or Follow-up required after 7 days.
Step 5: Find the Contact Page or Data Removal Request Form
Each website usually has at least one of the following pages: Contact, Contact Us, Privacy Policy, Data Removal, Opt-out (requesting the removal of information from the system), Legal, Support, or Help Center. Ask ChatGPT to find them for you instead of manually checking each page.
If you cannot find the contact page, try: checking the website’s footer, looking for the “Privacy Policy” page, searching for email addresses like [email protected], [email protected], or [email protected], or looking up domain ownership details using a WHOIS tool (a public lookup tool that shows who registered a website address).
Step 6: Draft the information removal request email
This is the most critical step. The email should maintain a clear, polite, and substantiated tone. Absolutely avoid using threatening, insulting, or overly emotional language. Doing so often leads websites to refuse cooperation rather than resolving the issue faster.
If you want to use it right away without waiting for ChatGPT to draft it, here is a pre-made email template; you just need to fill in the blanks:
Step 7: Draft a short message when the website only has a contact form
Some websites do not provide an email address, only a contact form on the page. In that case, use a shorter message to easily fit into smaller text boxes:
When a firmer tone with legal citations is needed
“Firm” does not mean abusive, threatening, or unprofessional. A strong request still requires all five elements: a specific link, specific information to be removed, a clear reason, a response deadline, and a request for confirmation after resolution—the only difference being the addition of legal grounds in Part 4 and a more assertive, non-pleading tone.
Step 8: Use ChatGPT agent if this feature is available on your account
As explained in Part 1, an agent is a mode that allows ChatGPT to independently open websites, find contact pages, and fill out forms on your behalf. This is a powerful feature, but it must be used with caution, because when logging into a website or enabling additional applications, the agent may access sensitive data such as emails, files, or account settings. Part 6 of the documentation provides detailed instructions on how to enable this mode, along with a dedicated set of prompts that allow the agent to handle almost the entire process.
The agent can assist with: opening each website in your tracking sheet, finding the contact or Privacy Policy page, filling out contact forms if you permit it, preparing removal request contents in advance, submitting the request when the website allows it and you confirm, and recording the status after submission.
What you should not let the agent do automatically: Do not allow the agent to automatically enter passwords for important accounts, submit identity documents without your prior review, agree to unfamiliar terms and conditions, make payments for takedown services, send mass requests without your approval of the content, or access your personal email and cloud storage unless absolutely necessary.
Step 9: Monitor Responses After Submitting the Request
Submitting does not mean your work is done. You need to record: the submission date, the target website, the email or form used, the content sent, whether there is a ticket ID (tracking number provided by the website), the follow-up deadline, the website’s response, and the final status.
Suggested timeline so you don’t forget: after three days, check if you received a confirmation email acknowledging receipt of the request. After seven business days, send a first reminder email if there is no response. After fourteen days, send a second reminder. After thirty days, consider next steps such as reporting to the platform, requesting search engines to remove outdated results, or seeking legal counsel if the information is overly sensitive.
Step 10: Request search engines to update or remove outdated results
In some cases, the original website may have already removed your information, but Google or other search engines continue to display a cached version or an outdated snippet. When this happens: verify whether the original link has truly been removed, take a screenshot if the search results still display your information, use the search engine’s outdated content removal or update tool, and check back after a few days.
Tip: Schedule periodic checks: If your account has access to the Tasks feature (automated reminders), ask ChatGPT to remind you to check periodically so you don’t have to remember manually. If you don’t have this feature, you can still schedule reminders on Google Calendar or on your phone.
6. Automate the entire process using ChatGPT agent
If you don’t want to manually complete each step in Part 5, you can let the agent mode handle almost everything: automatically finding the page, filling out the form, and submitting the application. You only need to confirm sensitive steps. Here is a 4-step process specifically for the agent, which can run in parallel with or replace Part 5.
Easiest way to enable agent mode
- Open the ChatGPT app or website, log in with a paid account: agent mode is currently only available for Plus, Pro, Team, or Enterprise plans, and is not yet available for free plans[2].
- In the message input box, click the plus icon (+) or the tools icon next to the chat box.
- In the list that appears, select “Agent mode” (or “Trợ lý hành động” if the interface is translated into Vietnamese).
- Start a completely new chat, paste the prompt from Step 1 below, and send it.
If your account does not show Agent mode yet, your app version may not be updated; try updating the app or using it on a desktop browser.
Step 1: Scan for data leaks using an agent
Prepare your full name, past and current phone numbers and email addresses, location, or field of work. Open a new chat (with Agent mode enabled), paste the following prompt, and send it.
Step 2: Draft the legal ultimatum
In the same chat thread, paste the following prompt to have ChatGPT draft a data takedown request based on the list just found.
Step 3: Activate the agent to process each site
Still in the same chat thread, paste the following prompt. The agent will sequentially open each link, automatically fill in the information, paste the letter content from step 2 into the reason field, and click the submit button.
Warning: The agent may pause and ask for your confirmation before executing sensitive actions, such as clicking the final “Submit” button on certain websites. This is a built-in safety mechanism in ChatGPT, not an error. Stay close to your screen while the agent runs to provide timely confirmation when prompted, rather than leaving it running unattended in the background.
Step 4: Set up monthly recurring sweeps
Data brokers are stubborn. Many sites automatically scrape your information again just months after removing it. This final prompt asks ChatGPT to use Memory (long-term memory) and Tasks (scheduling feature) to automatically repeat the process every month.
7. Quick-Use Prompt Pack
If you don’t want to re-read the entire process every time you use it, here are four condensed prompts for the manual workflow that pack all the necessary details. Just click copy and paste into ChatGPT.
If you want to let an agent handle almost the entire process instead of doing it manually using the four prompts above, use the set of four agent prompts in Section 6.
8. Checklist and Common Mistakes to Avoid
Step-by-Step Checklist
- Search for your name, email, phone number, and username on Google using quotation marks.
- Use ChatGPT to assist in finding related results and filter out those that actually pertain to you.
- Classify the sensitivity level of each result, then create a tracking spreadsheet.
- Locate the Contact, Privacy Policy, or Data Removal page for each website.
- Draft a removal request email, submit it via email or a contact form, and log the sent date and message content.
- Check for responses after three to seven days, and send a follow-up if it has not been processed.
- Once the website has removed the content, check whether Google still displays the outdated result; if so, request a search engine update.
- Schedule periodic checks weekly or monthly to ensure nothing is missed.
- Do not provide personal identification documents unless you are certain the website is trustworthy.
- Do not pay any unfamiliar data removal services without thoroughly verifying their credibility.
Six Most Common Mistakes
| Mistake | Why You Should Avoid It |
|---|---|
| Believing that ChatGPT crawls the entire internet | ChatGPT can search the public web, but it cannot guarantee finding every single page, database, or content hidden behind a private login. |
| Sending too much sensitive information into the chat | You should only provide enough information for searching and verification; do not send passwords, OTPs, bank account numbers, or identification documents unless absolutely necessary. |
| Sending overly emotional requests | Removal request emails should include specific links, a clear response deadline, and maintain a professional tone throughout the process. |
| Failing to follow up after submitting a request | Many websites will not respond immediately; you need a tracking sheet and should proactively send follow-up emails. |
| Confusing someone with the same name as yourself | If the results only match the name without clear evidence that it is you, mark it as “needs verification” instead of rushing to submit a removal request. |
| Allowing the agent to act too autonomously | If using an agent, require it to pause before submitting forms, entering sensitive information, or performing any critical actions. |
9. When to Seek Legal Support
In most cases, you can handle the issue yourself using the process above, based on Decree 13/2023/ND-CP mentioned in Section 4. However, consider consulting a lawyer or relevant authorities if you encounter any of the following situations:
- The exposed information includes photos of identity documents.
- Your home address, phone number, or family details are made public.
- Sensitive images are posted without permission.
- There is fabricated, defamatory, or threatening content targeting you.
- The website refuses to take down the content despite the information clearly being your private data, even after citing Decree 13/2023/ND-CP.
- There are signs of fraud, extortion, or impersonation.
- Customer data or company confidential data is leaked.
10. A Shortened Workflow for Busy People
If you find the full guide above too long, here is a condensed twelve-step version—just follow them in sequence:
- Open Google.
- Search for your name, email, and phone number, always enclosed in quotation marks.
- Copy suspicious links to a temporary storage space.
- Paste them into ChatGPT and ask it to categorize their sensitivity level.
- Ask ChatGPT to find the contact page for each website.
- Ask ChatGPT to draft a takedown request email, citing Decree 13/2023/ND-CP if a firmer tone is needed.
- Send the email or fill out the contact form, or let an agent do it for you as described in Section 6.
- Log the date sent in a simple spreadsheet.
- After seven days, if there is no response, ask ChatGPT to draft a follow-up email.
- Once the website removes the content, check back to see if Google is still displaying it.
- If it still appears, request the search engine to refresh its results.
- Schedule a recurring check every week or every month.
11. Troubleshooting tips when the agent gets stuck
Here are practical experiences to keep you from feeling overwhelmed when doing it yourself, even for the first time.
How to handle an agent getting stuck on a Captcha
When the agent encounters a Captcha while browsing the web, it will automatically pause and notify you on the screen. At this point, simply solve the Captcha manually directly on the screen displayed by ChatGPT, then type in the chat box: “I have solved the Captcha, please proceed by clicking the Submit button”, and the agent will continue working through the remaining pages.
Manual Cross-Check After One Week
About a week after the agent finishes submitting all requests, go to Google yourself and search for exact-match phrases by enclosing them in quotation marks, for example: "Nguyễn Văn A" "0901234567". If Google no longer returns results from the previous junk sites, your campaign was a success.
- Turned off “Improve the model for everyone” in Data Controls before entering sensitive information.
- Always run prompts belonging to the same workflow within a single chat session so the agent correctly remembers the context and website list.
- Stay near your computer while the agent is running to confirm emails or solve Captchas when prompted.
- Check back on Google after a week to verify the actual results instead of relying solely on the AI’s report.
12. Frequently Asked Questions (FAQ)
Below are detailed answers to the most common questions regarding automating the process of auditing and protecting personal data:

Chưa có bình luận nào. Hãy là người đầu tiên chia sẻ góc nhìn!