Skip to content

Username Search – Find Accounts on 3,000+ Sites

Try for free

Search one or many usernames across 7,000+ websites (social networks, coding, gaming, forums, dating) and get every matching profile URL. Maigret-grade detection rules with false-positive control.

🔎 Username Search – Find Accounts on 3,000+ Sites

Type a username and get every site where that handle has an account, with a direct profile link for each one. It checks social networks, developer platforms, gaming, video, music, dating sites and thousands of forums, using the open-source Maigret and Sherlock site databases (7,000+ sites in total). The top 500 sites take under a minute per username. No login, no API keys, and you only pay for accounts that are found.

✨ Why use Username Search?

  • ⚡ Fast — hundreds of sites are checked in parallel; the 500 most popular sites usually finish in under a minute.
  • 🎯 Fewer false positives — every run also checks a random username that cannot exist. Sites that claim to have that account too are dropped for the run, and you are not charged for them.
  • 🌍 Huge coverage — 7,000+ sites from the Maigret database (MIT) plus extra sites from Sherlock (MIT), ranked by popularity.
  • 🧾 Profile details on request — optionally save the public display name, bio, avatar, location, website and site IDs found on each profile.
  • 📋 Batch-friendly — paste a whole list of usernames; all results land in one dataset.
  • 💸 Pay per account found — no charge for sites where nothing was found or that could not be checked.

🎯 Use cases

  • 🕵️ OSINT investigations — map someone's public online footprint from one handle.
  • 🛡️ Trust & safety and fraud checks — see where a seller, scammer handle or suspicious account is also active.
  • 🏷️ Brand protection — find impersonation accounts and squatted handles that use your brand name.
  • 🧑‍💼 Due diligence — confirm that a counterparty's claimed profiles exist.
  • ✍️ Handle availability — check where a new brand or username is still free (turn on "Include not-found").
  • 📰 Journalism and research — link accounts across platforms for a story.

🚀 How to use

  1. Enter one or more usernames in Usernames (a leading @ or a pasted profile URL is cleaned up for you).
  2. Choose Number of sites to check (500 by default, 0 checks all 7,000+ sites).
  3. Optionally narrow by tags (e.g. social, coding, gaming) or list specific sites.
  4. Click Start and watch accounts appear in the Output tab.
  5. Export to JSON, CSV, Excel or HTML, or fetch the dataset through the API.

📥 Input

FieldDescriptionDefault
usernamesUsernames to search["nasa"]
topSitesCheck the N most popular sites (0 = all)500
tags / excludeTagsOnly / skip sites in these categories–
sitesOnly these sites, by name or domain–
falsePositiveCheckDrop sites that also "find" a random usernametrue
extractProfileFieldsSave public profile detailsfalse
searchRelatedUsernamesAlso search usernames linked from found profiles (max 10)false
includeNotFoundAlso save not-found and unconfirmed rows (free)false
includeSimilarSearchInclude sites that can only be checked via fuzzy search pagesfalse
requestTimeoutSecsWait per site10
maxConcurrencySites checked at the same time200
{
  "usernames": ["nasa", "spacex"],
  "topSites": 500,
  "tags": ["social", "video"],
  "extractProfileFields": true
}

📤 Output

One row per site where the username has an account.

FieldDescription
usernameThe searched username
siteNameSite name
profileUrlDirect link to the profile
siteMainUrlSite home page
statusClaimed (account exists); with "Include not-found" also Not Found, Unknown (blocked or timed out) or Unreliable (failed the false-positive check)
httpStatusHTTP status the site answered
isSimilarChecked through a search page that may match similar names
alexaRankSite popularity rank (lower is more popular)
tagsSite categories
detectionMethodstatus_code, message or response_url
errorTypeWhy a check could not be confirmed (only on Unknown rows)
foundViaFor related usernames: the profile that linked to it
fullName, bio, image, location, website, extractedIdsPublic profile details (only with extractProfileFields)
scrapedAtCheck time (UTC)

Example (fictional account):

{
  "username": "jane_doe_dev",
  "siteName": "GitHub",
  "profileUrl": "https://github.com/jane_doe_dev",
  "siteMainUrl": "https://www.github.com/",
  "status": "Claimed",
  "httpStatus": 200,
  "isSimilar": false,
  "alexaRank": 10,
  "tags": ["business", "coding", "networking"],
  "detectionMethod": "status_code",
  "errorType": null,
  "foundVia": null,
  "fullName": "Jane Doe",
  "bio": "Backend developer. Coffee and open source.",
  "image": "https://avatars.githubusercontent.com/u/0000000",
  "location": "Berlin, Germany",
  "extractedIds": {"uid": "0000000", "public_repos_count": "42"},
  "scrapedAt": "2026-10-14T18:00:00Z"
}

The run's key-value store also holds an OUTPUT summary: per username, how many sites were checked, how many profiles were found, how many checks could not be confirmed, and which sites were removed by the false-positive check.

💰 Pricing

You pay $0.01 per account found ($10 per 1,000 found accounts). Sites where nothing was found, unconfirmed checks and sites removed by the false-positive check are free. Platform usage (compute and proxy traffic) is billed separately and is small: one username on the top 500 sites uses about $0.04 of compute and proxy traffic.

Example: a username found on 150 of the top 500 sites costs about $1.50 in result fees plus about $0.04 of platform usage.

❓ FAQ

Is it legal? It only reads public pages that anyone can open without logging in. Profile details are personal data in many jurisdictions (GDPR, CCPA), so collect them only with a lawful basis, and do not use the results for decisions about employment, credit, housing or insurance.

Do I need a login or a proxy? No login. Datacenter proxies are used by default; switch to residential proxies to get through more sites that block datacenter IPs.

How accurate is it? Each site has its own detection rule (status code, page text or redirect). The built-in false-positive check removes sites that answer "found" for any name, and search-page matches are left out by default. Some popular sites (for example Instagram, LinkedIn, Reddit) often block automated checks; those come back as unconfirmed rather than as a wrong "not found".

Why are some fields empty? Profile details are only saved with extractProfileFields, and only for sites whose pages expose them. alexaRank is empty for sites without a popularity rank.

How many usernames can I search? Up to 1,000 per run. Each username is checked on every selected site.

Can I schedule it or use the API? Yes. Schedule runs to monitor a handle over time, or start runs and read the dataset through the API.

⚠️ Limits

  • Sites that require a login, solve a CAPTCHA or block automated traffic return Unknown instead of a result.
  • Site layouts change. The rules come from the community-maintained Maigret and Sherlock databases and a small share of them go stale between updates.
  • A found account means the handle is registered on that site, not that it belongs to the person you are looking for.
  • Checking all 7,000+ sites takes about 20–25 minutes per username and uses more proxy traffic (about $0.25 per username), mostly because many small forums are slow or offline.

Site data: Maigret (MIT, © Soxoj) and Sherlock (MIT, © Sherlock Project). Profile parsing: socid-extractor (MIT).

Related actors

G2
4.7/5
Excellent
Trustpilot

The internet is full of data.
Start owning yours.

Scrape.ly gives you the infrastructure, the tooling, and the team to extract, process, and operationalize web data at any scale. Setup takes minutes. The advantage lasts forever.

No credit card. No setup calls.