Fonts look harmless.

A designer picks a clean heading font. A theme ships an icon set. A developer adds Google Fonts because it’s fast to paste one line into the site. Nobody thinks of it as a business decision.

Then the website ships five weights, an icon font, a third-party font stylesheet, a late-loading hero headline, and a mobile layout that jumps when the real font finally appears.

That matters. Web fonts can affect page speed, visual stability, privacy, bandwidth, brand consistency, and whether a customer sees readable text quickly. The font file is small compared with a video, but it often sits close to the critical path. If it loads badly, people feel it.

These web font statistics are built for web professionals, business owners, and marketing teams who need a practical answer to a simple question: are our fonts helping the site, or making it heavier than it needs to be?

Web font adoption statistics

1. Web fonts appear on roughly 88% of websites. HTTP Archive’s 2025 Web Almanac found web fonts on about 88% of mobile pages, up slightly from about 87% in 2024. That means custom typography is no longer a design extra. It’s normal website infrastructure. (HTTP Archive)

2. Web font usage was only about 3% to 5% of sites in 2011. HTTP Archive’s font chapter says early web typography adoption sat in the low single digits when it started tracking the category. The jump from that range to 88% shows how quickly typography moved from novelty to default. (HTTP Archive)

3. Web fonts passed half of websites by 2015. The same HTTP Archive history says web fonts were used by over half of sites by 2015, then reached around 75% adoption by 2020. If your site still uses only system fonts, you’re now in the minority, but not automatically wrong. (HTTP Archive)

4. About 12% of websites still avoid web font files. Since HTTP Archive found web fonts on about 88% of sites, the remaining share is roughly 12%. For some businesses, system fonts are a smart performance choice. For others, it may just mean an old theme was never updated. (HTTP Archive)

5. Google Fonts still appears on about half of websites. HTTP Archive found Google Fonts on about 54% of desktop sites and 47% of mobile sites in 2025. Google remains the default font provider for a huge chunk of the web, even as more teams bring font files in-house. (HTTP Archive)

6. Google Fonts’ desktop reach fell from 57% in 2024 to 54% in 2025. That three-point drop does not mean Google Fonts is going away. It means self-hosting, privacy reviews, and performance work are slowly changing font delivery decisions. (HTTP Archive)

7. Adobe Fonts is used on about 4.2% of desktop sites and 3.5% of mobile sites. HTTP Archive calls Adobe Fonts much smaller than Google Fonts, but it still grew modestly in 2025. This is the brand-heavy end of the market: editorial sites, polished B2B brands, and companies with tighter visual systems. (HTTP Archive)

8. Font Awesome’s hosted service is down to roughly 3% to 4% of sites. HTTP Archive says Font Awesome was about 4.4% in 2022, about 4.0% in 2024, and roughly 3% to 4% in 2025. Many teams have moved icons to inline SVGs, sprite sheets, or local bundles. (HTTP Archive)

The business takeaway is simple: most sites use web fonts, but the good sites treat them like assets that need ownership. They don’t let every plugin, theme, and landing page builder add its own font stack forever.

Hosting and delivery statistics

9. About 72% of websites self-host at least one font file. HTTP Archive found that roughly 72% of websites serve at least one font from their own origin, either fully self-hosted or mixed with a font service. That is now the most widespread delivery pattern. (HTTP Archive)

10. Around one-third of sites are exclusively self-hosting fonts. HTTP Archive reports about 31.5% of desktop sites and 36.4% of mobile sites use only self-hosted fonts, with no third-party font service on the page. That is up from roughly 30% the year before. (HTTP Archive)

11. Mixed font hosting is still common. About 36.1% of desktop sites and 31.7% of mobile sites use a combination of self-hosted fonts and an external font service. In real audits, this often shows up as a theme font, a Google Font, and an icon kit all loading together. (HTTP Archive)

12. Google Fonts plus self-hosted fonts appears on about 36% of desktop sites and 32% of mobile sites. This hybrid setup is one of the most common combinations HTTP Archive found. It’s also a frequent source of duplicate typefaces and unnecessary requests. (HTTP Archive)

13. Google Fonts only appears on about 13% of desktop sites and 12% of mobile sites. These sites rely exclusively on Google’s font CDN for custom fonts. That setup can be easy to manage, but it still adds a third-party dependency to the rendering path. (HTTP Archive)

14. The top three hosting patterns account for roughly four out of five websites. HTTP Archive says self-hosted only, Google only, and Google plus self-hosted are the dominant configurations, while other combinations fall below 1% quickly. That makes font audits more predictable than most website cleanup work. (HTTP Archive)

15. Cache partitioning reduced the old CDN advantage for shared Google Fonts files. HTTP Archive notes that browser cache partitioning removed some of the historical performance benefit of relying on Google’s CDN because fonts are not broadly shared across unrelated sites the way developers once assumed. (HTTP Archive)

16. Locally served Font Awesome files account for about 8.5% of desktop font requests and 9.3% of mobile font requests. HTTP Archive found self-hosted Font Awesome among the most frequently fetched font resources on the web. If your site uses only three icons, loading a full icon font deserves a hard look. (HTTP Archive)

For a small business, self-hosting is not automatically better. It only works when someone owns the setup: WOFF2 files, correct MIME types, caching, preloading where needed, and no stale font files left behind by old themes.

File format and payload statistics

17. WOFF2 accounts for about 65% of font file requests. HTTP Archive found WOFF2 at 65.2% of desktop font requests and 65.4% of mobile font requests. It is now the standard format for web fonts because it compresses well and is broadly supported. (HTTP Archive)

18. WOFF still represents about 16% of font requests. HTTP Archive found older WOFF 1.0 at 16.0% on desktop and 15.8% on mobile. Some of that is legacy fallback. Some of it is just old font files nobody cleaned up. (HTTP Archive)

19. WOFF and WOFF2 together account for about 81% of all font requests. HTTP Archive says the web has largely consolidated around those two formats. If a normal business site is still shipping EOT, SVG fonts, or raw old files for broad public traffic, that’s usually a cleanup opportunity. (HTTP Archive)

20. TTF still shows up in about 6.7% to 6.8% of font requests. HTTP Archive found TTF at 6.7% on desktop and 6.8% on mobile. Raw TrueType files are often a sign of a copied font package that was never converted for the web. (HTTP Archive)

21. Around 6.6% to 6.7% of font requests are served as application/octet-stream. HTTP Archive says this usually means the server is sending the wrong MIME type, such as serving a WOFF2 file without font/woff2. That is not a design problem. It’s a server configuration problem. (HTTP Archive)

22. The median font file is about 35 KB to 36 KB. HTTP Archive’s 2025 data shows typical font transfer sizes in the mid-30 KB range across desktop and mobile. One good font file is not the problem. Too many variants, old formats, and unused icon sets are the problem. (HTTP Archive)

23. Three-quarters of web fonts are under about 80 KB. HTTP Archive reports the 75th percentile in the mid-70 KB range. That is manageable, but it still adds up when a page loads regular, bold, italic, semi-bold, headings, icons, and a separate campaign font. (HTTP Archive)

24. The 90th percentile font file is roughly 115 KB to 116 KB. HTTP Archive says the largest 10% of font files are well above 100 KB, often because of complex scripts, variable fonts with many axes, or un-subsetted files. (HTTP Archive)

25. The largest 1% of font files are still hundreds of kilobytes. HTTP Archive says the heavy tail often includes CJK fonts, icon fonts with embedded graphics, or fonts shipped with extra debug tables. This is where subsetting can save real bandwidth. (HTTP Archive)

26. In 2025, self-hosted mobile fonts were roughly 74% WOFF2, while Google-hosted mobile fonts were roughly 77% WOFF2. HTTP Archive’s service breakdown shows self-hosted fonts are catching up, but Google still tends to serve optimized formats more consistently. (HTTP Archive)

This is the cleanup checklist I would use on most business sites:

  • Keep only the weights and styles the site actually uses.
  • Prefer WOFF2, with older fallbacks only when there’s a real browser support reason.
  • Replace large icon fonts with SVGs when the site only needs a small icon set.
  • Subset large font families, especially multilingual fonts.
  • Check server MIME types and long-term caching headers.

Loading behavior and Core Web Vitals statistics

27. About half of pages using web fonts specify font-display: swap. HTTP Archive found font-display: swap on 49.6% of desktop pages and 50.1% of mobile pages in 2025. That means many teams now prefer visible fallback text instead of waiting with invisible text. (HTTP Archive)

28. font-display: block appears on about one-quarter of pages. HTTP Archive found block on 24.7% of desktop pages and 24.9% of mobile pages. The key detail: about 70% of real-world block usage comes from icon font styles, not body text. (HTTP Archive)

29. font-display: optional appears on only about 0.4% to 0.5% of pages. HTTP Archive describes optional as the most aggressive performance choice because the browser may keep the fallback font rather than swapping late. It is still rare. (HTTP Archive)

30. Font preconnect is used on 18.3% of desktop pages and 17.7% of mobile pages. HTTP Archive found preconnect is the most popular font resource hint. It can help when a page depends on a third-party font origin, but it should be used carefully because unnecessary early connections are also waste. (HTTP Archive)

31. Font preload appears on 12.0% of desktop pages and 11.7% of mobile pages. HTTP Archive says preload usage has risen, which makes sense for critical above-the-fold text. Google’s web.dev guidance also explains that preloading a known font URL can start the font request earlier instead of waiting for CSSOM and render tree discovery. (HTTP Archive, web.dev)

32. Google considers LCP good at 2.5 seconds, INP good at 200 milliseconds, and CLS good at 0.1 or less. HTTP Archive’s 2025 performance chapter uses those Core Web Vitals thresholds. Fonts can affect the user’s path to those scores when text is late, layout shifts after a font swap, or critical rendering is blocked by font loading decisions. (HTTP Archive)

33. Only 48% of mobile sites had good Core Web Vitals in 2025. HTTP Archive reports mobile Core Web Vitals improved from 36% in 2023 to 44% in 2024 and 48% in 2025, while desktop reached 56%. Fonts are not the whole problem, but they are one of the small decisions that can push a borderline site the wrong way. (HTTP Archive)

The practical rule: don’t preload every font. Preload the one or two critical files that determine above-the-fold text, then let the rest behave like non-critical assets.

Variable font and advanced typography statistics

34. Variable fonts are used on 39.4% of desktop sites and 41.3% of mobile sites. HTTP Archive says about four in ten sites now use at least one variable font, up from roughly 33% of desktop sites and 34% of mobile sites in 2024. (HTTP Archive)

35. Noto Sans JP is the most used variable font. HTTP Archive says Noto Sans JP appears on roughly 25% to 26% of pages using variable fonts. That matters because global language support is a major driver of modern font usage, not just brand styling. (HTTP Archive)

36. Noto Sans JP, Roboto, Open Sans, and Montserrat account for almost 60% of variable font requests. HTTP Archive found variable font usage is top-heavy, with a few open and widely available families carrying a large share of real-world adoption. (HTTP Archive)

37. Weight is the most-used variable font axis. HTTP Archive found the wght axis explicitly adjusted on about 57% to 61% of pages using variable fonts. Most teams are using variable fonts as multi-weight files, not as experimental typography systems. (HTTP Archive)

38. Variable font animation appears on only about 0.3% of all pages. HTTP Archive says even among pages that use variable fonts, only about 0.7% to 0.8% animate or transition variable font axes. That is probably healthy. Animated typography can look good, but it can also create layout shifts and distraction. (HTTP Archive)

39. Color fonts appear on only about 0.05% to 0.06% of websites. HTTP Archive says that is roughly one in 2,000 pages. Color fonts have grown from a tiny base, but for most business websites, SVGs and normal font files still cover the real needs. (HTTP Archive)

What this means for business websites

A font audit is not about making the site ugly. It’s about making sure the typography earns its keep.

The sites that usually need attention have the same pattern: Google Fonts from the theme, a second font from the page builder, an icon font from a plugin, a separate landing page font, and six weights when the design uses two. Each item feels small. Together, they create drag.

A good font setup for a normal business website is usually boring:

  • One primary text family, maybe one display family.
  • Two or three weights, not every weight the family offers.
  • WOFF2 files served from the right place with the right MIME type.
  • font-display: swap or a deliberate alternative for text.
  • Preload only for critical above-the-fold font files.
  • SVG icons instead of a full icon font when the icon set is small.
  • A fallback font stack that does not cause ugly jumps.

If your site is running slowly, fonts probably are not the only problem. Images, JavaScript, hosting, third-party scripts, and CMS bloat often matter more. But fonts are a clean win because the fixes are specific. Remove unused weights. Convert to WOFF2. Stop loading duplicate families. Replace the giant icon font. Set display behavior. Test the page again.

That’s not glamorous work. It is the kind of work that makes a site feel tighter.

If you want help finding the hidden weight in your website, start here: get a practical website review from Your Web Team. We’ll look at the parts customers actually feel: speed, clarity, leads, forms, tracking, and the technical clutter that slows everything down.

FAQ

Are web fonts bad for performance?

Not by themselves. HTTP Archive found the median font file is only about 35 KB to 36 KB, but performance problems show up when a site loads too many families, too many weights, old formats, duplicate providers, or late critical fonts. (HTTP Archive)

Should I self-host Google Fonts?

Sometimes. Self-hosting can reduce third-party dependency and give you more control, but only if the files are optimized, cached, served as WOFF2, and maintained. HTTP Archive found about 72% of sites already self-host at least one font file, so the practice is normal, but execution still matters. (HTTP Archive)

What is the best web font format in 2026?

For normal public websites, WOFF2 is the default choice. HTTP Archive found WOFF2 accounts for about 65% of font file requests, far ahead of other formats. (HTTP Archive)

How many font weights should a business website use?

Most small business sites are fine with two or three weights. For example, regular, medium or semi-bold, and bold. More weights can be useful for a design system, but every extra file needs a reason.

Do fonts affect Core Web Vitals?

They can. Fonts can contribute to late text rendering, layout shifts, and slower perceived loading. HTTP Archive’s 2025 performance chapter defines good Core Web Vitals as LCP within 2.5 seconds, INP within 200 milliseconds, and CLS at 0.1 or less. (HTTP Archive)