We Checked 60 Usernames Across 3,166 Sites: Here's What We Found

How many websites already have an account under the username you are about to pick? Every username tool, ours included, can show you the answer for one name. We wanted the bigger picture, so on October 13, 2026 we ran 60 test usernames through our own checker against every site in our list: 3,166 sites, 189,960 checks in total.

The short version: in our October 2026 test, a random made-up handle already existed on about 7 sites on average (median 3), once we removed five sites that answer "found" for any name. Common names such as johndoe existed on 866 sites on average. Length mattered more than anything else. None of our 15 long random handles showed up anywhere real.

A username collision is when the handle you want is already registered by someone else on a site. This study measures how often that happens for different kinds of handles. All figures below come from our own run and are in the summary CSV.

Key findings

• Common names are taken almost everywhere. Our 10 generic handles (johndoe, janedoe, alex, sam, mike, emma, chris, test, user123, gamer) existed on 866 sites on average, about 27% of everything we checked. The median was 780 and the range ran from 340 to 1,473.
• Random handles are mostly free. Across 50 made-up handles, the average was 11.9 sites, or 6.9 after removing five catch-all sites. The median was 3.
• Length is the biggest factor. Six-character handles like kett70 existed on 10.8 real sites on average, 8 to 10 character handles on 9.1, and 13 to 18 character handles on zero.
• The big platforms fill up first. On our 20 most popular platforms, common names were taken on 11 of them on average, short random handles on 5, medium handles on 4 and long ones on none.
• About one check in seven fails. 8.8% of checks came back Blocked and 5.1% Unknown or error. Six of the 20 big platforms gave no clear answer for any of our 60 handles.

Common names vs random handles

Common names were taken on far more sites than random handles: 861 on average against about 11 or fewer.

Handle group	Handles	Avg. real sites where it exists	Of top 20 platforms taken (avg.)
Common names	10	861	11.1
Short random (6 chars)	15	10.8	5.2
Medium random (8-10 chars)	20	9.1	4.0
Long random (13-18 chars)	15	0	0

The gap is enormous. A handle like alex or test has been registered on hundreds of forums, code hosts and small communities over the past twenty years. A handle made of a word and a few digits has usually never been registered anywhere.

The chart barely shows the random handles, which is the point. If you want a name that is free nearly everywhere, avoiding dictionary words and first names on their own does most of the work. Our guide to choosing a unique username covers how to build one you can still remember.

How length changes the picture

Longer handles were taken on fewer sites. Within the random handles, shorter ones were taken more often, and the spread was wide. The busiest six-character handle existed on 23 real sites; the quietest on 2. Medium handles ranged from 1 to 33, with a median of just 3.5, so most were free almost everywhere and a few happened to collide with real accounts.

The platform names that fill up fastest are the ones with huge user bases and short, free-form usernames. Roblox showed 31 of our 50 random handles as already in use, including all 15 short ones, and TikTok showed 24. GitHub, by contrast, had only 4 of the 50, and Steam none.

What happens when you check 189,960 usernames

Most checks end in Not found, as you would expect for mostly random names. The part worth noticing is the 14% that end in Blocked or Unknown. A tool that shows only "taken" or "available" has to put those somewhere, and if it calls them available, you get a wrong answer about one time in seven. That is why our results keep Blocked and Unknown separate.

Which kinds of sites block checks

Blocking was very uneven. Shopping sites blocked 34.2% of our checks, hobby sites 23.6% and gaming sites 20.2%. Video sites blocked almost nothing (0.1%).

Speed varied less. Among categories with at least 20 sites, shopping sites answered fastest (0.28 seconds on average) and social sites close behind (0.39 seconds), while blogs were slowest at 0.94 seconds. A handful of archived and legacy sites averaged 1.7 seconds. Small categories had the most unclear answers: about a quarter of checks on health and dating sites came back Unknown.

The big platforms: what we could and could not see

For the 20 platforms with their own checker pages, the results split into three groups.

• Clear answers every time or nearly: TikTok, YouTube, Discord, Roblox, Twitch, GitHub, Snapchat, Steam, Bluesky, Telegram and Chess.com. X gave a clear answer for 48 of 60.
• No clear answer at all: Instagram and Spotify (all Unknown), and Reddit, Xbox, PlayStation and Etsy (all Blocked). For these, our result means "check by hand".
• One-sided: Pinterest never returned Not found, so free names show as Unknown. Threads never returned Found, not even for common names, so its Not found is weak evidence.

Each platform's checker page now has a dated "What we tested" box with its own numbers, starting from the platform username checkers hub.

False positives we found and checked by hand

Our long random handles, such as hollowpebble1650, almost certainly have no accounts anywhere. Five sites still reported them as Found every time. We opened those pages by hand:

• Three sites (a self-hosted code forge, a link aggregator and a video site) returned a "Making sure you're not a bot!" challenge page with HTTP status 200. Their detection rules only look at the status code, so the challenge page counts as a profile.
• One community forum returns the same app page for any name.
• One music community returned "not found" to us by hand but a success code to our server, so the server is seeing a different page.

We excluded these five from the "excluding catch-all" figures, and we have flagged their rules for fixing. We also spot-checked Found results the other way. Repeating Roblox's username check by hand, it said a 6-character random handle was already in use and accepted our long handle as valid. A GitHub account and a Telegram page for two random handles both turned out to be real. Those are other people's accounts, so we are not naming them.

Methods

• When and where: October 13 to 14, 2026, from one server, using the same public API and detection rules as our username search.
• Sites: all 3,166 sites in our list that day, in 19 categories. No sampling.
• Handles: 10 common generic handles we chose, plus 50 random handles from a seeded generator using nature and object words with digits: 15 short (6 characters), 20 medium (8 to 10) and 15 long (13 to 18). No real people's names.
• Pace: 40 sites per request, at most 4 requests at a time, with a pause between requests. We never hit our own rate limit.
• What we kept: only counts by status, handle group, category and site. The checker reads whether a page exists; we stored no profile data.
• Catch-all rule: a site that reported Found for at least 80% of the long random handles is treated as answering for any name. Five sites met it.

We keep the full aggregates and a longer methods note; ask at [email protected] if you want them. We plan to rerun the study to see how the numbers move.

Download the summary data (CSV): run details, status shares, results by group, the five catch-all sites and all 60 test handles. This run was on 13 October 2026, on the 3,166-site list we used then; we moved to our own engine on 21 October, and today’s list has more sites, so these figures describe that run only. A Found result here means a public profile answered, not that the name is unavailable to register.

Limitations

• This is one day from one network location. Sites change their pages and blocking constantly, and a rerun will differ.
• Found means a public page answered for that exact name. It does not prove an active account, and the catch-all filter will not catch every false positive.
• Not found can be wrong too. Private accounts, reserved names and hidden deactivated accounts can look free. Always confirm on the platform's sign-up page.
• Our random handles come from a small word list, so they show what happens to a typical made-up name, not to real usernames people pick. The generic names were chosen by us.
• Response times were measured inside our checker and include each site's own delay, not what you wait in the browser.

What this means if you are choosing a username

Pick something that is not a first name or a single word, and make it at least 10 to 12 characters if you want it free nearly everywhere. Check it once across every site with the name checker, then claim the platforms you care about the same day. For sites that come back Blocked or Unknown, open the sign-up page and try the name there.

Worked example. Say you want nightowl92. It is a word plus two digits, 10 characters, so it sits in our "medium" group, where the median was 3.5 sites. Type it into the username search, which checks every site in our list from your request. On a name like this, Found results are most likely other people. Then look at Blocked and Unknown and try those sites' sign-up pages by hand. Instagram and Threads are not checked live, so use their manual guides. If too many big platforms are taken, try a longer variant such as nightowlreads92. In our run, handles of 13 or more characters were free on every real site.

If you are checking your own digital footprint instead, the numbers cut the other way: a Found result for a short or common handle is very likely someone else. Our digital footprint check explains how to tell your accounts apart.

You are welcome to quote these figures. Please link to this page as the source.