What Do Newly Registered Domains Run? Our Methodology

The question, stated carefully

"What do new websites run?" sounds simple and hides three separate questions. Which domains count as new? Which of them host a real site? And how do you learn what a site runs without crawling the web yourself? Each has a defensible answer, and the answers shape every chart. This page explains TLDlive's choices so that you can judge the results.

No percentages appear on this page on purpose. The statistics are live and the mix changes daily, so a number printed here would be stale quickly. The current leaderboard, trends and per-extension breakdowns are on /technologies.

Step 1: defining a new domain

TLDlive builds its lists of added and dropped domains from daily ICANN CZDS zone files, for the registries that participate. A domain is "new" on the day it first appears in a zone snapshot that it was absent from the day before. The mechanics, and their limits, are explained in zone files and CZDS and in the guide to newly registered domains.

Two caveats matter here. First, appearing in a zone is not the same as being registered: a domain can be registered days before it receives name servers. Second, some country-code registries do not publish zone files, so they are absent from the added-domain analysis. Per-extension tables therefore cover participating registries only.

Step 2: where the technology data comes from

TLDlive does not crawl newly registered domains to fingerprint them. The technology statistics come from observations contributed by people using the TLDlive Lens extension who chose to share them. When a participant visits a site, the extension detects technologies locally and, with consent, sends a small domain-level record: the registrable domain, detected technology names, a purpose category and a few booleans. The exact fields are listed in the privacy policy. Those records are joined to the zone-file data, so that a domain can be placed in the "added in the last N days" group.

This design has consequences that are worth stating plainly:

• The sample is people's browsing, not the whole registry. A domain appears only if some participant visited it. Domains nobody opens, which includes much of the unused, parked and abusive part of any registry, are under-represented.
• Popular destinations are over-represented. A new domain that gets links or advertising is visited more, so it is more likely to be observed.
• The participant base is not random. Lens users skew toward developers, marketers and domain professionals, and their browsing reflects that.
• Counts are of domains, not visits. A domain counts once per day per install; a heavily visited site does not outweigh a quiet one.

So the correct reading of a chart is "among observed new domains, share X ran technology Y", not "X percent of all new domains run Y". The size of the observed sample is shown beside the charts so you can judge how much weight a figure deserves.

Step 3: what counts as a technology

Detection uses a catalogue of fingerprints: patterns in response headers, meta tags, script and stylesheet URLs, a capped sample of the HTML, cookie names, a few JavaScript globals and DOM selectors. Each match has a confidence. A name enters the statistics only if it exists in the extension's catalogue, so unknown software never appears. Categories such as content management system, ecommerce, CDN, analytics and web framework come from the same catalogue. How individual detections work is explained for specific cases in what CMS a website uses, Cloudflare and CDN detection, and analytics and tag managers.

Detection can miss things. Server-side tools leave no trace in the browser, and hardened sites strip version strings. A missing technology means "not detected", never "not used".

Step 4: separating real sites from parked ones

A large share of any new-registration list is parked or unused. A parking page runs its own technologies, and counting it as a "website" would distort the leaderboard toward parking platforms and their CDNs. Lens therefore classifies each page's purpose, with parked as one category, using public signals such as parking-page patterns. The statistics can then be filtered by purpose, so that a chart for ecommerce sites does not include parking pages. The method is described in detecting parked domains.

Reading the analytics

View	What it shows	How to read it
Leaderboard	Technologies ranked by distinct domains observed	Compare shares within one period and one extension, not across different sample sizes
Per technology page	Share, trend and sample domains for one technology	Check the number of domains behind the trend line before drawing conclusions
By extension	Technology mix for one TLD	Differences between extensions often reflect who registers there, not platform quality
By purpose	Mix of ecommerce, blog, SaaS, parked and other categories	Purpose is a heuristic guess; treat small differences as noise
Rising technologies	Technologies whose share is growing	Small bases make large percentage swings; look for steady growth over weeks

What the data can and cannot support

Reasonable uses: spotting which platforms are growing among new sites, comparing the technology profile of two extensions, finding candidates for outreach such as agencies targeting a particular platform's users, and checking how a security tool's population differs from the average site.

Unreasonable uses: estimating the absolute market share of a platform, judging an individual domain's quality from a single observation, or treating the numbers as a census. When the sample for a specific cell is small, the page should say so, and you should discount accordingly.

Reproducing the idea yourself

If you want to run a similar study, the recipe is: obtain zone file diffs for the added domains, resolve which have web servers, fetch the home page with a polite, identified crawler that respects robots.txt, run a fingerprint catalogue on the response, and store only derived labels. TLDlive uses the browser-based approach because it sees real rendered pages, including JavaScript globals, and because opt-in observations avoid crawling the sites ourselves. Each approach has bias: crawlers over-sample unvisited and abusive sites, browser panels over-sample popular ones. Report which you used.

Sources

• ICANN Centralized Zone Data Service: czds.icann.org
• ICANN, zone file access policy (CZDS terms): icann.org
• W3Techs, methodology of web technology surveys, for a crawler-based comparison: w3techs.com

Frequently asked questions

What percentage of new domains run WordPress?

Check the live leaderboard on the technologies page. The figure changes over time and differs by extension, so TLDlive does not print a fixed number in this guide. Read it as the share among observed new domains, not among all registrations.

Does TLDlive crawl newly registered domains?

No. Technology data comes from opt-in observations by Lens users, joined to the zone-file data about when domains were added. Domains that no participant has visited are not in the technology statistics.

Why are parked domains separated out?

Parking pages run their own technologies and make up a large part of many new registrations. Counting them as websites would skew the platform rankings, so the purpose category lets you filter them out.

Can I get the underlying data?

Aggregated statistics are published on the technologies pages and through the API described on the developers page. Raw events are retained for a limited period and are not published.