Name validation API: catch junk names, not real people

An assessment, a score and the reasons

Each name comes back as plausible, suspicious or implausible, with a score from 0 to 100 and the list of signals behind it. “asdf qwerty” is implausible with keyboard_pattern on both words; “Mickey Mouse” is suspicious as a fictional character; “Jennifer Null” is plausible, and the score goes up because Jennifer appears in official birth records more than a million times.

That last part is what our name data adds. Most validators only know what a bad name looks like. NameGender also knows what a real first name looks like, from counted birth and population registers in eleven countries.

curl -X POST https://namegender.com/api/v1/name-check \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"name": "asdf qwerty"}'

# "assessment": "implausible",
# "score": 0,
# "signals": [
#   { "code": "keyboard_pattern", "severity": "high", "part": "first_name", "value": "asdf" },
#   { "code": "keyboard_pattern", "severity": "high", "part": "last_name", "value": "qwerty" }
# ]

What it catches

High-severity signals make a name implausible on their own; medium ones make it suspicious, as do two low ones.

Real names are left alone

Name filters have a long record of blocking real people: Jennifer Null cannot book flights, Cockburns trip profanity filters, and people with a single name cannot complete forms. The rules here are built around those cases. Null, Test and Doe are fine as surnames; only “null” on its own or the pair John Doe is flagged. Mickey Smith passes, Mickey Mouse does not. Profanity matches whole words only, and a first name with counted birth records is never treated as one.

Single names, two-letter names such as Ng, short words without vowels such as Strč, and names in scripts our data does not cover are not penalised, and a first name we have never seen does not count against anyone on its own. We tested the rules on 4,610 real first names in nine scripts, alone and paired with a surname: none was marked suspicious or implausible.

Flag, don’t reject

Use implausible to hold a sign-up for a second look or ask for the name again, and suspicious to mark a lead for review. Rejecting people automatically on a name is the one use we advise against: the cost of a wrongly blocked customer is higher than the cost of one junk lead.

The check works in the same places as the rest of the API. In Google Sheets, =NAMEGENDER_CHECK(A2:A500) marks a whole column. In Zapier, the Check Name action returns the signal codes as plain text, so a filter step can route anything containing profanity or placeholder. Make, n8n and all ten client libraries have it too.

Questions

Related pages