ThreatPort

Is WordPress secure? We measured 2,759 live websites

First-party measurement of 444 WordPress sites against 2315 non-WordPress sites drawn from the same sample. WordPress came out ahead on more checks than it lost, with one clear exception: 39.4% publish a DMARC record and leave it switched off, against 26%.

2026-08-05 · 3,000 domains drawn · 2,759 answered · 444 WordPress · margin of error ±5.1 points on any difference

The short answer. WordPress sites were not worse. They averaged 72.2 out of 100 against 66 for everything else in the same sample — a real gap, not noise (95% CI 4 to 8.5). Of nine checks, 4 favoured WordPress and 1 favoured everything else, with 4 too close to call. The one place WordPress clearly lost is the one that matters most for impersonation: 39.4% publish a DMARC record and leave it at p=none, which reports abuse and blocks nothing, against 26% of other sites.

What was compared

CheckWordPressEverything elseDifference
No DMARC record at all 25.9% 35.3% -9.4
DMARC published but left at p=none 39.4% 26% +13.4
DMARC set to quarantine or reject 34.7% 38.6% —
No SPF record 15.5% 21.9% -6.4
No TLS on the apex 1.6% 16.2% -14.6
No HSTS header 57% 66% -9
No Content-Security-Policy header 78.6% 77.3% —
Passes the header check 28.8% 29.9% —
Passes all four checks 12.8% 16.2% —

A dash means the gap is inside the margin of error and should be read as no difference. With 444 WordPress sites and 2315 others, any gap under 5.1 points is noise.

The finding that matters

WordPress sites are more likely than other sites to have set up DMARC — only 25.9% publish no record at all, against 35.3%. They are also far more likely to have left it switched off. Put together, roughly the same share of each group is actually protected, but on WordPress the failure mode is different: the record exists, someone configured it, and it stops nothing.

That is worse than having no record, in one specific way. A missing DMARC record is a gap somebody can still discover. A record at p=none looks like a completed task.

Version numbers in public

35.4% of the WordPress sites measured print their exact version in the page's generator tag. That is not a vulnerability by itself, and this study did not check whether any of those versions were current. It is simply information the site gives to everyone who asks.

How this was measured

Domains ranked in the Majestic Million top 100,000 by referring subnets, of which 3,000 were drawn with a seeded Fisher-Yates shuffle (mulberry32, seed 20260805). Each domain is fetched once, as a browser would fetch it, and classified as WordPress or not from what that single page already publishes about itself. Majestic ranks by inbound links rather than traffic, which favours content sites — where WordPress is most common — so the WordPress share of this sample is not a estimate of its share of the whole web.

Each domain was fetched once over HTTPS, as a browser would fetch it, and its DNS TXT records, MX records and TLS certificate were read. Nothing else was requested: no port scanning, no subdomain enumeration, no login pages, no plugin or vulnerability probing, and no path other than the home page. The page markup was read to classify the platform and then discarded. No individual site is named, here or in the published data.

What this does not show

This is a measurement of configuration, not of exploitability. WordPress's security reputation rests largely on third-party plugins and themes, and none of that is visible from outside. A site can pass every check on this page and be running an abandoned plugin with a public exploit. Read the result narrowly: on the things the whole internet can see, WordPress sites are configured about as well as everything else.

The comparison group is also not a clean control. "Everything else" includes API endpoints, CDNs and infrastructure hosts that were never meant to serve a web page, which is the likeliest reason the TLS gap is as wide as it is — 16.2% of them answered on the apex without a certificate, against 1.6% of WordPress sites, which are almost all on managed hosting where a certificate is provisioned automatically.

The sample is drawn from a link-based ranking, which favours content sites — where WordPress is most common. The WordPress share of this sample is therefore not an estimate of its share of the web. The comparison between the two groups inside the sample is unaffected, which is why the comparison is the finding and the share is not.

Check your own domain

The four checks on this page are the ones ThreatPort runs on any domain, free and without an account.

Run a free external scan

Raw aggregates: data.json · CC BY 4.0, attribution to ThreatPort