CSS frameworks in 2026: Bootstrap, Tailwind and the large remainder

On 27 August 2026 we parsed the home pages of 399 reachable hosts, looking for unambiguous CSS framework fingerprints in the markup:

The first thing that stands out is not the framework shares but the size of the remainder: about two thirds of sites in every zone show no sign of any common framework. Surveys of "CSS framework market share" usually compute percentages inside that other third and stay quiet about it being a third.

Bulma appeared on 3 sites in .ru and 2 in .org, UIkit on one in .ru, Materialize and Foundation on single sites. Shares cannot be computed at those counts, so we give absolutes.

The clearest difference is Tailwind: 3% in .ru against 13% in .com and 12% in .net. Bootstrap, conversely, is most common in .ru at 28%.

A caveat this figure requires: three Tailwind sites in the .ru zone is far too few to state a precise share. Only the direction is solid here, and it is supported by Tailwind appearing 2.5–4 times more often across all three international zones.

An explanation consistent with our other measurements: Tailwind presupposes a build step, where classes are generated and pruned at publish time. That fits a "framework plus deploy" workflow and fits poorly with a site edited directly in CMS templates. Given that the .ru zone is led by WordPress, Tilda and Bitrix, the low Tailwind share looks like a consequence of how sites are built rather than of front-end taste.

Detecting a framework from markup is a method with known flaws, and it is more honest to say so up front.

Fingerprints overlap. Utility classes such as px-4 exist in both Tailwind and Bootstrap 5. Our first detector tripped on exactly this — it put Tailwind in .ru at 5% instead of 3%. The final version relies only on unambiguous signals: numbered colour scales (bg-slate-800) and responsive prefixes (md:flex) for Tailwind; grid classes col-md-6 and filenames for Bootstrap. Sites matching both came to 1–2 per zone.

A build step erases traces. Tailwind discards unused classes at publish time, and Bootstrap is often pulled through a preprocessor with renaming. A site can use a framework and show no recognisable sign — those fall into the "neither" column.

One page is checked. Home pages are frequently built separately from the rest of a site.

All of this skews the result one way: real framework use is higher than measured. So comparing zones is legitimate, while taking our shares as an absolute estimate is not.

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