Headless CMS in 2026: why the share cannot be measured from outside

On 27 August 2026 we parsed the markup of 401 home pages (150 hosts per zone, drawn from those users brought to our tools) and looked for content management system fingerprints.

Traditional systems:

Headless systems, across all four zones combined:

That is 11 pages of 401 — 2.7%, against roughly a third for traditional systems.

Here is the substance, and it matters more than the figures.

A headless system leaves no trace in markup — that is its whole point. A traditional CMS signs itself: WordPress leaves /wp-content/ in paths and a generator meta tag, Bitrix leaves /bitrix/. A headless system serves content over an interface while a front-end framework assembles the page. Nothing in the resulting HTML points to the source.

Our seven Contentful sites were found only because images load from its own delivery network and the address is visible in the markup. A project caching those images itself would become invisible.

Hence the honest conclusion: 2.7% is not the share of headless systems but the share of those that happened to leave a trace. The true figure is higher, and by how much cannot be determined from outside.

The bias is moreover asymmetric: we see nearly all traditional systems and nearly no headless ones. So comparing "2.7% against 30%" is invalid — it compares not two shares but the detection completeness of two different classes.

Our method is not worse than the accepted one — it is the same one. Major CMS market surveys work exactly this way: crawl pages and look for fingerprints. So the limitation is common to all of them, and every published headless share is understated for the same reason.

The Web Almanac acknowledges it outright: detection tools identify a technology by a recognisable trace, and "unless these tools leave detectible fingerprints, we are unable to measure their usage". That statement applies to every such measurement, ours included.

There is a second source of bias that gets discussed less. The home page is what gets checked. Large projects often keep the front page on a static generator while using a headless system for a blog or a documentation section. Such a site enters the statistics as "no CMS".

What follows practically: treat any figure of the form "headless CMS share has grown to N%" as a lower bound with an unknown distance to the truth. It shows direction; it does not show magnitude.

Despite all of the above, part of the measurement holds — the part concerning traditional systems, which are detected dependably.

In the .ru zone Bitrix is more common than WordPress — 22% against 11%. That reverses the international zones, where WordPress leads by a wide margin and Bitrix appears in single figures. Our cut agrees in direction with the W3Techs data, though our denominators differ and the numbers are not directly comparable.

Tilda appears only in .ru — eight pages against one or two across every other zone combined. For a builder aimed at one market that is expected, and the data confirms it.

Drupal and Joomla have all but vanished — one or two pages per zone. Systems that held a visible share ten years ago sit at the edge of visibility in our sample.

To see what is detected on a specific site, use the technology detector.