How to search common names without mixing people

When several people share a name, the job is not to build one fuller profile. It is to keep candidates separate until independent evidence connects the same person across sources. Start with the full name and one known public anchor, review candidates individually, and stop a merge when dates, locations, employers, handles, or photos conflict.

Define the person before you search

Write down only what you already know from a legitimate source:

• full name and known spelling variants
• approximate city or country
• current or former employer, school, profession, or public role
• a known username, website, company domain, or profile URL
• a relevant date range

Mark uncertain details as uncertain. A guess such as “probably lives in London” should not become a filter that hides the correct candidate or a fact that gets repeated later.

Start with the strongest anchor

Use an exact username, official profile URL, company-domain email, or uncommon name plus context before using a bare common name. If all you have is the name, run the name search and treat every result as a separate hypothesis.

The DeepSearch transparency statement explains that names alone do not establish a match. This is an intentional tradeoff: an incomplete result is safer and more useful than a detailed profile assembled from several namesakes.

Compare candidates in a fixed worksheet

Use the same fields for every candidate rather than following whichever result looks most convincing first.

| Candidate | Stable handle or URL | Location | Work or school | Time fit | Contradictions | Source links | | --- | --- | --- | --- | --- | --- | --- | | A | | | | | | | | B | | | | | | | | C | | | | | | |

This reduces confirmation bias and makes a later correction possible.

Require more than one kind of agreement

Useful identity links include:

• one public profile directly linking to another
• the same distinctive handle plus matching biography details
• an official employer page matching a public professional profile
• a personal domain linking to the account
• consistent location and work history across independent sources
• the same image with a traceable history and other matching context

Two pages copying the same directory record are not necessarily independent. Neither are two search snippets that point back to one source.

Weak signals include a shared first and last name, a generic job title, a common city, a similar headshot, or a username that many people could choose. Several weak signals can help prioritize review, but they do not automatically become one strong signal.

Treat contradictions as information

Do not average away a conflict. Keep candidates separate when you find:

• overlapping full-time roles in different countries
• incompatible ages or graduation dates
• a known profile cross-linking to a different handle
• an image whose earlier source identifies another person
• different middle names or legal entities
• activity timelines that cannot both describe the same person

A contradiction may mean one source is stale, but that needs evidence. Until then, label the records as different or unresolved.

Use hints to rank, not to prove

DeepSearch may accept context such as an approximate location or known handle to help order candidates. Those hints come from the searcher and can be wrong. A candidate appearing first does not mean the hint was verified, and a match percentage is not the probability that every field in the profile is true.

Open the source links for the fields that matter. The terms describe match scores as estimates and require verification before acting on a result.

Pivot one anchor at a time

Once a candidate has a verified anchor, start a separate lookup from that anchor:

1. use the exact handle in username search
2. use a publicly listed company address in email search only if you already have the full address
3. use a publicly listed number in phone search only if it is relevant to the task
4. reverse-search a public profile image to trace where it appeared, not to identify the face

Keep the new result attached to the anchor that produced it. Do not pour every new field back into all same-name candidates.

Know when to stop

Stop and label the identity unresolved when:

• only the name agrees
• the strongest source is a copied aggregator page
• important dates or locations conflict
• the task would require private-account access, password-reset testing, impersonation, or contacting unrelated people
• resolving the identity would create a safety or harassment risk

“Possible match” is a valid outcome. The absence of enough evidence is not a reason to lower the standard.

For a general records workflow, use how to search public records without confusing people with the same name. For account-level checks, continue with how to verify a social media account.