Where do you start when you have almost nothing?
With whatever you have, ranked by how much it reaches, which is rarely the order people expect.
An email address reaches 71 sources here and a handle reaches 71 platforms. A name reaches 13 registers and a phone number reaches 8. So a name and a number, which feel like the most personal things you could hold, are the weakest places to begin.
If all you have is a name, the first job is not to search harder. It is to convert what you have into something a service can answer a yes-or-no question about, because that is the only kind of question the open web reliably answers.
A workplace, a city, an old handle, a photograph. Any of those turns a name from a dead end into a starting point.
How does one identifier become another?
Because results are not just answers. They are new identifiers, and that is what makes this compounding rather than linear.
A service confirming an address often exposes a display name beside it. A profile page carries an avatar and sometimes a bio with a second handle in it. A public commit history carries the address that authored it. Each of those is a fresh input, and running it reaches platforms that were never going to answer the question you started with.
Two or three hops is where most real findings live. The first search tells you somebody exists; the second tells you who they are; the third is usually where you either confirm it or discover you have drifted.
Past that, returns fall off sharply and the risk of drifting onto a stranger rises with every hop.
How do you know you are still following the same person?
You do not, by default. That is the part nobody writes down, and it is where confident wrong conclusions come from.
Every pivot is a hypothesis. The handle on this profile is the same handle as the one in that address is a guess, and a reasonable one, and sometimes wrong. Handles get recycled after an account is deleted, squatted deliberately, and picked independently by strangers who happened to like the same word.
The discipline is to hold the hypothesis loosely and look for something that could disprove it. A join date years before the person was plausibly online. A location that contradicts everything else. An avatar that is a stock photograph.
Hunting the contradiction beats counting the matches, because matches accumulate whether you are right or wrong.
Following a trail is the easy part. Whether the next print belongs to the same foot is the question this whole method turns on.
What counts as enough corroboration?
Two sources that did not get it from each other, and that qualifier does most of the work.
Half the people-search industry resells the same broker files. Three services agreeing on somebody's middle initial is not three confirmations; it is one file, quoted three times, and the agreement is manufactured. Independence means the sources had separate reasons to know the fact.
What genuinely counts: a platform confirming its own account exists, a register holding its own filing, a breach corpus containing its own dump, a page the person wrote themselves. Those know things first-hand. Anything that summarises other sources is a convenience, not a witness.
Prefer one first-hand source to five copies. The question to ask of any result is how this source came to know. If the answer is that it read it somewhere else, it corroborates nothing, however confident it looks.
When does a pivot go wrong?
Four ways, and three of them look exactly like success while they are happening.
- A common handle. A real word as a username belongs to dozens of unrelated people, and every one of them returns a hit.
- A recycled account. The handle is the same string and the person behind it changed years ago.
- A shared address. Household, business and legacy addresses are used by more than one human being.
- Your own momentum. After two confirming hops, the third gets read generously. This is the one that actually catches people.
The defence against all four is the same: write down which hop was a guess, at the moment you make it, rather than reconstructing your confidence afterwards.
A note costs nothing now and is the only thing that lets you unwind a conclusion later without redoing all of it.
When should you stop?
When the next hop would not change what you do, which is earlier than curiosity suggests.
Most searches have a real question underneath them: is this person real, is this account theirs, has this address been around. Once that is answered the rest is accumulation, and accumulation is where a reasonable check turns into something that would be uncomfortable to explain.
There is a practical version of the same rule. If you cannot say what a finding would change, you are collecting rather than investigating, and the collection has a cost: everything gathered is something you now hold about somebody who did not ask you to.
That cost is also yours. A pile of unexplained findings is hard to justify afterwards, and afterwards is when justification gets asked for.
What does this look like on a real report?
Messier than a diagram, and the mess is the informative part.
A run measured in July 2026 against one address returned 19 confirmed accounts and 60 breach records out of the 74 sources checked then, in 24 seconds. Eleven sources errored and four were rate-limited: a fifth of the run never answered at all. None of that is unusual and all of it changes how the result should be read.
Which is why every source is listed here with what it returned, including the failures. A source that answered and found nothing is evidence about the person. A source that timed out is evidence about the internet. A report that shows you only the hits has quietly merged the two, and the merge is where somebody concludes an account does not exist when it simply was not checked.
What are you not allowed to do with this?
More than people assume, and the restrictions are on use rather than on searching.
In the United States, using any of this to decide somebody's employment, housing, credit, insurance or tenancy puts you under the Fair Credit Reporting Act, and a consumer lookup service is not a consumer reporting agency. That use is not permitted here. In the EU and the UK, public personal data is still personal data, so processing it needs a lawful basis under the GDPR.
Repeated contact is where this becomes a crime. Finding somebody lawfully says nothing about whether approaching them is welcome. Sustained unwanted contact is harassment in most jurisdictions no matter how the details were obtained, and the search history will be read as intent.
How to run this on DetectiveCheck
Three habits rather than three buttons, because the tool is the easy part.
- Start with your strongest identifier, not your most personal one. Address or handle before name or number.
- Run whatever the first report hands back. A display name or a profile URL is a new identifier and usually the productive one.
- Mark every hop you could not confirm, while you are making it. Findings here carry their source and a confidence rating so this stays visible on the page.
The failures are listed beside the hits for the same reason: a source that timed out is not a source that said no, and only one of those is evidence about the person.
You can see the whole shape in the sample report without creating an account.
Common questions
How do you find information on someone online?
Start from the identifier that reaches furthest and pivot. An email address reaches 71 sources and a handle reaches 71 platforms, against 13 for a name and 8 for a phone number. Each search returns new identifiers — a display name, a profile URL — and running those reaches places that would never have answered your original question.
What is pivoting in OSINT?
Using a result as the input to the next search. A service confirming an email address often exposes a display name beside it; that name is a handle you can run, and the handle returns profiles the address never would. Two or three hops is where most real findings live, and where most mistakes are made.
How do I know I found the right person?
By looking for something that would disprove it rather than counting matches. A join date earlier than the person could plausibly have been online, a location contradicting everything else, a stock photograph as an avatar. Matches accumulate whether your hypothesis is right or wrong, so only contradictions are informative.
What counts as corroboration?
Two sources that did not get the fact from each other. Half this industry resells the same broker files, so three services agreeing can be one file quoted three times. Genuine corroboration comes from sources with first-hand reasons to know: a platform confirming its own account, a register holding its own filing, a page the person wrote.
Why do I keep finding the wrong person?
Usually a common handle, a recycled account, or a shared address. A real word as a username belongs to dozens of unrelated people and every one returns a hit. The fourth cause is momentum: after two confirming hops the third gets read generously, which is the failure that catches experienced people.
Is it legal to find information on someone?
Searching publicly available sources is legal in the United States and the EU. Use is restricted: in the US, deciding employment, housing, credit, insurance or tenancy from a lookup falls under the Fair Credit Reporting Act and is not permitted with a consumer service. In the EU and UK, public personal data still needs a lawful basis under the GDPR.
When should I stop searching?
When the next finding would not change what you do. Most searches have a real question underneath — is this person real, is this account theirs — and once it is answered the rest is accumulation. Everything gathered beyond that point is something you hold about somebody who did not ask you to.
Does the person know I am researching them?
No. These searches query public sources and the account-existence behaviour platforms expose to anybody. Nothing is sent to them. What does reach them is what you do next: opening a profile on a network that shows visitors, or making contact, are visible actions the search itself never takes.
