Your name, your old street address, and the email you abandoned years ago are probably still on data broker sites right now — packaged, indexed, and for sale to anyone with a credit card. A prompt shared by the Telegram channel @gptpublic (211 views, 17 reposts, 6 reactions when it went out) turns that mess into a to-do list: hand an AI agent your details, and it hunts across data brokers, people-search sites, directories, social networks, and forums — then hands back a table you can use to delete digital footprint records one by one. The post's headline promises a five-minute sweep that erases the whole trail at once. The honest version is less flashy — and a lot more useful.
What happened
The post is a copy-paste prompt — a block of instructions you drop into an AI agent that can browse the web. Think of it as a job description for a very patient intern. No coding, no subscription, no app to install.
The prompt, in plain English
You hand the agent your raw identity data: every spelling of your name, your age, the country and city you live in now, the places you used to live, your usernames, your email addresses, and links to profiles you've already confirmed are yours.
Then you give it marching orders. Start with people-search sites and directories in the countries where you've lived, then branch out to other sites that fit. Search names, usernames, and combinations of those with locations.
What comes back is a table with five columns: the site, a status ("yes," "maybe," or "couldn't verify"), a link to the record itself, the reason the agent thinks it's you, and the site's official opt-out page.
Two tools it leans on
The prompt also asks the agent to run Social Analyzer locally, following the tool's README (the plain-text instruction file that ships with open-source software) to check your usernames across dozens of platforms. If it can't get the tool running, you're told to say so and keep searching other ways — a nice touch, since most prompts would quietly skip the step and pretend.
The second tool is Have I Been Pwned, a free site where you type in an email address and it tells you which known data breaches included it. The prompt is firm here: check whether you show up. Don't go requesting the leaked data itself.
The rules that keep it honest
A few guardrails separate a useful list from a pile of false alarms:
• Never guess a link. The agent has to point at the page it actually visited.
• "Yes" is reserved for matches where the name *and* the location line up, with nothing contradicting your other details. A username hit on its own can't be better than "maybe."
• No buying reports, no creating accounts, no firing off removal requests on its own.
• Every check gets a date stamp, so you know how stale the list is.
The caveat that matters
Independent coverage of AI-assisted removal tools makes the same point. Agents are genuinely good at finding your data and drafting removal requests, but they're not a magic eraser. Mika Reyes and La Sorsa both frame this as an ongoing loop: find, verify, request, then re-check in a few months, because brokers re-list records constantly. A common name or a broker that ignores opt-outs still needs a human in the loop.
What it means for you
You don't have to be a privacy obsessive for this to matter. Here's where it actually lands in normal life.
At home
A stranger can pull your old addresses and your relatives' names in about three clicks. That's how a lot of scam calls open — the caller already knows your maiden name and the last two streets you lived on. Getting those listings taken down makes you a noticeably harder target for phone and mail fraud.
At work
Recruiters, clients, and hiring managers Google you before they ever reply to your email. If an outdated listing shows an old job title next to your current city, that's the version of you they see first. Cleaning it up is cheap reputation maintenance.
For your business
If you're a solo founder or run a one-person shop, your home address is often your business address in public records. That means junk mail, door-knocks, and a paper trail you'd rather not advertise. The table the agent builds doubles as a starting inventory.
For students and early-career folks
Club rosters, old forum posts, and dead portfolios from a decade ago are still crawlable. This is a clean way to find the embarrassing corners of your own history before an admissions officer or a recruiter does.
For creators
Usernames are the whole problem. You may have used the same handle across dozens of sites you've long forgotten. Social Analyzer was built for exactly that job — and the prompt's rule that a username-only match caps at "maybe" saves you from chasing strangers who happen to share your handle.
For your income
Freelancers and sellers get impersonated constantly, because a fake profile needs nothing but a name and a photo. Knowing where your real listings live makes the fakes easier to spot and report.
If you want to get comfortable with AI helpers before you start feeding one your personal details, MyKreaTool collects free AI tools you can experiment with safely first.
How to try it right now
Step 1: Check your email for free (about 2 minutes)
Open Have I Been Pwned, type your main address, hit search. No account, no install, nothing to pay. You'll get a list of breach names — nothing more, which is exactly the point.
Step 2: Paste the prompt into an AI agent
Use any assistant that can browse the web. Fill in the bracketed fields with your real details:
```
Find public records about me on data brokers, people-search sites,
directories, social networks, and forums.
My data:
• Name and spellings: [name]
• Age: [age]
• Currently live in: [country, city]
• Previously lived in: [countries, cities]
• Usernames: [list]
• Emails: [list]
• Confirmed profiles: [links]
Start with people-search and directory sites in the countries I've lived,
then find other relevant sites. Search by names, usernames, and
combinations of those with locations.
Also run Social Analyzer locally per its README for username searches.
Re-check every match on the actual page — a tool's score doesn't confirm
identity. If you can't run it, say so and keep searching other ways.
Give me a table: site, status (yes / maybe / couldn't verify), link to the
record, basis for the match, official opt-out. Only mark "yes" when the
name and location match with no contradictions. A username alone is
"maybe" at most. List sites with no results and unreachable sites
separately, with the reason.
For every "yes" or "maybe," build a removal checklist: exact page,
required fields, steps, and whether email, phone, or document
confirmation is needed.
Also check whether my email appears in known breaches via Have I Been
Pwned. Don't request the leaked data itself.
Don't guess links, don't buy reports, don't create accounts, and don't
send requests. State the date you ran the check.
```
Step 3: Run Social Analyzer yourself
It's a free, open-source tool that runs on your own machine, so your usernames never leave your laptop. Follow the README to install it, then treat its findings as leads, not verdicts.
Step 4: Work the table from the top
Tackle the "yes" rows first, then the "maybe" rows. Each opt-out form is slightly different — some want nothing but an email address, some make you confirm through a link they send you, and a handful still ask for a phone call or a photo ID.
Step 5: Set a reminder for 90 days
Records come back. Brokers rebuild their databases from public sources, so the same listing can reappear. A calendar reminder and a re-run of the same prompt is the whole maintenance plan.
Step 6: If you'd rather not do it by hand
Automated services exist for exactly this. Optery handles broker removals on an ongoing basis, and MindStudio has a walkthrough of Unbroker Hermes, an agent built for the same job. Try the free route first — you'll learn which sites actually hold your data.
Upsides and what changes
The biggest win isn't the deletions. It's the inventory. Most people have no idea how many listings carry their name, and once you've seen that table you can't unsee it. You'll also spot patterns — a username you reused everywhere, an email that's been in several breaches, an old address that's still attached to your current phone number.
Second, the prompt does the tedious part. Writing opt-out requests by hand for twenty sites is a weekend. Turning that into a checklist you work through in twenty-minute blocks is realistic.
Third, the honesty rules make the output trustworthy. Insisting on a real page link and capping username matches at "maybe" means you won't waste an afternoon filing a removal request for somebody with your same name in Ohio.
Limitations
This is not a delete button, and anyone selling it as one is overselling. The agent can find records and draft requests, but it can't push some forms through — several brokers deliberately require a phone call, a notarized letter, or a photo ID, and no amount of automation skips that. It also can't remove anything you've posted yourself: your old tweets, your public Facebook posts, the forum thread you started years ago are yours to delete, one by one, and a few sites will simply refuse. Public records tied to property, court filings, or business registrations are usually permanent in some form; you can sometimes suppress how easily they're found, not erase them. And because brokers rebuild from public sources, anything removed today can quietly reappear next quarter. Complex cases — a common name, a name change, a stalker situation — should go to a privacy lawyer or a dedicated service, not a chatbot.
Conclusion and one action for today
AI agents won't vaporize your digital footprint in five minutes, but they will hand you a map of it — and that's the part that used to cost hundreds of dollars and weeks of digging. The people-search sites, the directories, the usernames you forgot, the breaches your email is sitting in: all of it becomes a list you can actually work through. One action for today: open Have I Been Pwned and check your main email address. It takes two minutes, it costs nothing, and it tells you whether the rest of this is a curiosity or a priority.
Comments 0