23,303 potential licence issues, 22,289 websites, 150 foundries: what one verified pass of the live web found
Abstract. We scanned the live web for trial and demo builds of commercial fonts, then re-downloaded every candidate, read its internal name table and checked it actually draws text in a browser. One verified pass turned up 23,303 potential licence issues across 22,289 distinct websites and 150 foundries, which tells us trial fonts in production are not a rare slip but a quiet, everyday habit of the web.
Key findings
- 23,303 potential licence issues on 22,289 distinct websites, spread across 150 foundries.
- 95% of cases (22,030) carry an explicit trial marker inside the font's name table ("Trial", "Demo", "Unlicensed"). The other 5% (1,273) are flagged by foundry trial file naming.
- Roughly 1 in 3 candidates dropped out at the live re-check, so every case here was still serving a trial font on the day we checked.
- ~1,600 cases (about 1 in 15) sit on top-1M websites. This is not just hobby blogs.
- The top 10 foundries account for ~60% of cases; the median foundry has ~42 leads.
- 1,160 cases (~5%) are variable fonts.
- At list price, the unlicensed use adds up to at least ~€4.7m in licences.
Wait, what's a trial font doing on a live site?
Foundries are generous. Most let you download a trial build so you can mock up a logo, test a headline, or pitch a client before paying. Those files are usually clearly labelled: the font's own internal name says "Trial" or "Demo", and some trials are cut down in other ways too.
They're meant for the sketchbook. Not the shop window.
And yet. That gorgeous display serif on your local coffee shop's site? There's a real chance it's a trial file someone forgot to swap out. We wanted to know how real that chance is, so we went and counted.
The headline number
One verified pass found 23,303 potential licence issues on 22,289 distinct websites. That's a lot of websites. The median site serves 2 trial styles, and about 1 in 10 serves 7 or more. Seven trial styles is not an accident in a corner of a footer. That's a whole type system running on a free sample.
We say "potential licence issue" on purpose, and we'll keep saying it. A trial file on a live site is strong evidence, not a verdict. Plenty of site owners have actually paid for the licence and simply uploaded the wrong file (more on that in a sec). Context matters.
How sure are we? Pretty sure, and here's why
Anyone can grep the web for the word "trial". That gets you noise. Here's what each case had to survive:
- Web-scale scan to find candidate sites loading suspicious font files.
- Live re-check: we download the font the site is serving today. About a third of candidates fall out here, because sites change, fonts get licensed, pages die.
- Trial marker: we read the font's name table for "Trial", "Demo" or "Unlicensed". Where that's missing, we fall back to the foundry's own trial file naming, which we treat as weaker evidence.
- In-browser render check: does the font actually draw visible text? A file sitting unused on a server doesn't count.
- Dated evidence: file hash plus screenshot, so anyone can check our work.
Dev and staging sites are filtered out. Duplicates are counted once.
| Evidence type | Cases | Share |
|---|---|---|
| Name-table marker ("Trial", "Demo", "Unlicensed") | 22,030 | 95% |
| Foundry trial filename only | 1,273 | 5% |
| Total | 23,303 | 100% |
That 95% figure is the one we're proudest of. In the vast majority of cases the font file is literally telling you it's a trial.
Big sites, small sites, everyone's at it
About 1,600 cases, roughly 1 in 15, are on websites ranked in the top 1 million. So this is not only abandoned portfolios and weekend side projects. Real businesses with real traffic are shipping trial fonts.
That makes sense when you think about how sites get built. Big or small, the font usually gets chosen once, early, by whoever's designing. Then the site launches, the team moves on, and nobody ever checks the font folder again.
Where the cases cluster by foundry
The spread across foundries is lumpy, which is roughly what you'd expect if the most-loved typefaces get trialled the most.
| Leads per foundry | Number of foundries |
|---|---|
| 1,000+ | 8 |
| 100 to 999 | 24 |
| 10 to 99 | 118 |
| Total | 150 |
The top 10 foundries carry about 60% of all cases. But the median foundry still has around 42 leads, which, for a small studio living on licence sales, is not nothing.
Which kinds of sites?
About half the cases could be confidently classified by industry. Within those, the leaders were:
- Software and AI: 16%
- Food and drink: 13%
- Health and beauty: 11%
- Education and non-profit: 10%
- Design studios: 10%
Yes, design studios. We'll dig into that (gently) in our industry piece.
Variable fonts are in the mix
1,160 cases (about 5%) are variable fonts. One variable file can cover an entire family's worth of weights, so a single trial variable font in production can stand in for what would otherwise be a multi-style licence.
Does this match what others see?
Our pass lines up with broader census-style counts. In one snapshot, trial-marked fonts appeared on roughly 22,700 hosts. Looking at homepages that self-host a given foundry's fonts, somewhere between 5% and 16% were serving trial builds. And it's sticky: in an earlier pass, around 13,000 new candidates were re-checked and about two thirds were still serving the trial.
So this isn't a one-off spike. Trial fonts stay put once they land.
What it's worth
Priced at list, using the cheapest licence that would cover each case at the smallest web tier, the total comes to at least ~€4.7m. That's a floor, not a ceiling, and our separate valuation paper explains exactly why.
So what should anyone do with this?
If you're a designer or site owner: open your font folder. If a file name or font name says Trial or Demo, swap it for the licensed build (you may already own it!) or buy the licence. It's a five-minute fix.
If you're a foundry: this is revenue you've already earned the right to. The people using your work clearly love it. Most will happily pay once someone shows them, kindly, what's going on.
Methodology
- Window: data checked 13 to 15 November 2026, single FontRights verified pass.
- Pipeline: web-scale scan, live re-download of the served font, trial marker read from the font name table ("Trial", "Demo", "Unlicensed") or foundry trial file naming (weaker signal), in-browser render check for visible text, dated evidence (file hash and screenshot).
- Exclusions: dev and staging environments filtered; duplicates counted once; around a third of candidates removed at live re-check.
- Industry: classified for roughly half of cases; percentages are of the classifiable set.
- Language: every case is a potential licence issue. We don't name sites, fonts or foundries publicly, and we never contact site owners without a foundry's written authorisation.
- How to cite: FontRights (2026). The State of Trial Fonts on the Web 2026. Software Automated Ltd. Data checked 13-15 November 2026.
Run a foundry and curious what this looks like for your catalogue? There's a free locked dashboard preview that shows your counts, evidence mix and list value, with sites masked, before you commit to anything. Say hi at max@fontrights.com or have a look at https://fontrights.com.
Published by FontRights (https://fontrights.com), revenue recovery for type foundries. Contact: max@fontrights.com
Source: https://fontrights.com


