Born Yesterday: 20.1% of new domains have a real website by the end of day one
Every day dns.pizza diffs the zone files of about a thousand gTLDs against the day before and knocks on the door of every new name. Over the last 14 days that was 1,960,055 births and 1,617,867 deaths. Here is what a domain's first day looks like.
The daily rhythm
Births per day, newest on the right. Each bar opens that day's full census. The window's busiest day was Friday, August 21, 2026 with 181,687 new names.
What is behind a brand-new domain
Most new names are not websites yet. Across the window 30.9% answered nothing at all and 16.6% resolved but served no site; only 20.1% had a page with real content by the end of the day they appeared. Parked names are a smaller share than folklore suggests: 2.4%.
What the live ones are about
Subjects come from a language model reading the page text; “unclassified” means it was not confident enough to say. Hosting and domain businesses, gambling and retail lead the classified share almost every day.
Where they land
Who parks them
What live sites are built with
Where the trouble concentrates
Every birth is scored on its name, its infrastructure and its first-day page. Over 7 scored days, 5,291 of 2,321,244 births scored high or critical. A model looked at the top 2,100 and called 1,009 of them phishing, scam or abuse. Those are opinions, not findings; the named rows, with their evidence, are on the flagged list.
TLDs with the most flagged births per 1,000 (500+ births in the window)
Brands most echoed in new names
Signals that fire most
Nameserver providers of flagged names
Registrars of flagged names
Method
Zone files come from ICANN's Centralized Zone Data Service for every gTLD that publishes one, plus a sample of .com and .net under Verisign's zone-file programme. A birth is a name in today's zone and not yesterday's; a death is the reverse. Diffing needs no DNS queries. Each birth is then probed once over HTTP from a host that identifies itself and honours robots.txt, and classified by what came back. The risk scorer is deterministic and additive over public word lists; its weights ride along in the API. A language model judges the top few hundred rows a day from a screenshot and page text, and a second, stronger model gives a second opinion on the positives.
Counts include every zone. Names in .com and .net are not listed by name until the redistribution terms of that zone's access agreement are confirmed. Everything on this page is free as JSON at /api/zone/*; the daily hub is /newborn.