Font foundries have offered free trial downloads for years. The idea is sensible. A designer tries a typeface in a real layout, likes it, and buys a licence. The trial file is limited, clearly labelled, and meant to be thrown away.

The question we set out to answer was simple. How often does that file never get thrown away, and ends up serving text to real visitors on a live website instead?

The headline numbers

In our most recent verified pass, we re-checked every candidate site live between 13 and 15 November 2026. A site only counted if it was still serving a trial or demo build of a commercial font on that day.

23,908potential licence issues, verified live
22,858distinct websites
150foundries affected
1,160cases serving a variable font

The gap between the first two numbers is small but real. Some websites serve trial builds from more than one foundry, so they appear once for each. A site built from a template can carry two or three trial families at once.

For context, a wider census of fonts on home pages found trial-marked fonts on roughly 22,700 hosts in a single snapshot. Across individual foundries, trial builds showed up on somewhere between 5% and 16% of the home pages that self-host that foundry's typefaces. It is not a fringe problem. For many foundries it is a steady slice of all their web use.

Why "verified live" matters

Raw scan results go stale quickly. Websites are redesigned, files get swapped, businesses close. In one earlier pass, we re-checked about 13,000 newly found candidates. Roughly two in three were still serving the trial font. The other third had changed or disappeared.

That is why a scan alone is not a lead. A lead is a site we loaded again, whose font file we downloaded again, and whose trial marker we read again from inside the file. Anything older than that is a rumour.

How the evidence breaks down

Of the 23,908 cases, 22,559 (94%) carry the trial marker inside the font's own name table. That is the strongest kind of evidence. The foundry itself wrote "Trial" or "Demo" into the file, and the site is serving that exact file.

The remaining 1,273 (5%) match a foundry's own trial file naming. That is a weaker signal, so we grade it lower. It is still useful, especially for foundries whose trial files are otherwise identical to the retail ones.

A long tail of foundries

The distribution across foundries is steep. A handful of foundries account for most of the cases, and a long tail has only a few each.

Foundries by number of leads
Leads per foundryFoundries
1,000 or more8
100 to 99924
10 to 99118
Total150

The ten foundries with the most cases account for about 60% of the total. The median foundry has about 42. That spread probably tracks how foundries hand out trials. A full-family trial that downloads in two clicks travels much further than one sent by email on request.

It also means the problem is not limited to big names. Small independent foundries are well represented in the long tail. For a studio that sells a few dozen web licences a year, ten unlicensed sites is a meaningful number.

What kind of sites are they?

Mostly small ones. About 1,600 cases, roughly one in fifteen, are on sites inside the top million websites by traffic. The rest are the long tail of the web: studios, restaurants, shops, practices, start-ups and event sites.

About half of all cases can be placed in an industry. Of those, software and AI companies are the largest group, at around 16%. Food and drink (13%), health and beauty (11%), education and non-profits (10%) and design studios (10%) come next. Health and beauty, retail, fashion, education and finance made up most of the rest. A small number were large organisations.

None of this is surprising. Trial fonts end up where budgets are tight and sites are built quickly. That does not make the licence any less owed. It does shape how the follow-up should sound.

Variable fonts and family size

About 5% of all cases involve a variable font. That share varies a lot by foundry: for eight foundries it is above a fifth, and for one it is over a third. Variable fonts matter for value. One variable file can stand in for a whole family, and many foundries price it that way.

Most sites serve a small number of styles. The median is two. About one case in ten serves seven styles or more, which usually means a full family has gone live.

What the numbers do not say

Every case here is a potential licence issue, not a legal finding. We cannot see a foundry's sales records. Some of these sites will have bought a licence and uploaded the trial file by mistake. That happens a lot. The designer had both files in the same folder, and picked the wrong one.

A foundry is the only party that can tell the difference. It matches the leads against its own customers. What remains after that match is the real gap.

The numbers also undercount. We only count sites where we can read a clear trial marker, and many foundries ship trials without one. A trial build with a clean name table is invisible to this method. So is a site that only uses the trial font on inner pages.

The short version

  • More than 23,000 live cases of trial fonts on websites, across 150 foundries, in a single verified pass.
  • 94% of them carry the trial marker inside the font file itself.
  • Most are small businesses. The problem is spread across hundreds of foundries, large and small.
  • Each case is a starting point for a licence conversation, not a conclusion.

Want to see the cases for your own typefaces? Every foundry can get a free locked preview of its dashboard. Email max@fontrights.com.

Read next