A reader once emailed me after I published a piece with tight 1.3 line-height and a 90-character line length, telling me she'd given up two paragraphs in because the text physically hurt to track across the screen. She has dyslexia. I'd optimized that layout for how the page looked in a screenshot, not for how it actually got read.
Dyslexia affects roughly 20 percent of the population (Yale Center for Dyslexia and Creativity), which means a chunk of every audience you write for is fighting your typography before they've read a single sentence. Most sites clear the WCAG contrast minimum and call it accessible. That's the floor, not the ceiling.
TL;DR: Dyslexia affects about 1 in 5 people. Lexend, a Google-backed variable font, improved reading speed by 19.8% over Times New Roman in controlled testing (Google Design). Font choice helps, but line length and spacing matter more than most teams assume.
Which Fonts Actually Have Research Behind Them?
Fewer than the marketing copy around "dyslexia fonts" would have you believe. Lexend is the strongest evidence-backed option right now, developed with reading-speed research and funded through Google Fonts, and it works as a drop-in variable font with no licensing cost. OpenDyslexic gets recommended constantly, but the research is genuinely mixed, one PMC-published study found no measurable improvement over Arial or Helvetica, and readers in preference tests often picked the standard fonts instead. Atkinson Hyperlegible, designed by the Braille Institute, targets low-vision readers specifically but overlaps well with dyslexia-friendly traits: wide apertures and letterforms built to be told apart.
My actual recommendation, having tested all three across client projects: default to Lexend or a clean system sans like Inter, and if budget allows, add a font-switcher so readers can opt into OpenDyslexic themselves. Don't force one typeface as "the" dyslexia solution. There isn't one.
| Font | Research backing | Best use case |
|---|---|---|
| Lexend | 19.8% faster reading speed vs Times New Roman in controlled testing (Google Design) | Default body font, free variable font |
| OpenDyslexic | No measurable improvement over Arial/Helvetica in PMC-published study; readers often preferred standard fonts | Optional font-switcher choice only, not a default |
| Atkinson Hyperlegible | Designed for low-vision readers by the Braille Institute; overlaps with dyslexia-friendly traits | Sites already prioritizing low-vision accessibility |
What Do These Three Fonts Look Like Next to Each Other?
The frame above sets the same sentence in all three, plus whatever sans your own device serves as a control, at 16px with the spacing from the next section already applied. Read the bdpq line before the paragraph. Those four letters are the mirror pairs that get cited in every dyslexia thread, and they're where the three fonts genuinely disagree.
OpenDyslexic weights the bottom of every letterform and tilts several of them, which is the whole idea: gravity at the baseline is meant to stop a b flipping into a d. It also costs you something nobody mentions. I measured that same 79-character sentence in Chromium at 16px, and it runs 947px wide in OpenDyslexic against 572px in Liberation Sans, the metric clone of Arial my Linux test box serves. That's 66 percent wider. Set your measure in ch and swap the font, and the column physically grows by two thirds; set it in pixels instead and your 65-character line quietly becomes a 40-character one.
Lexend is the quiet one. Same sentence, 625px, about 9 percent wider than the Arial clone. Its one real flaw, and you can see it in the frame, is that capital I and lowercase l are the same bare vertical stroke. Atkinson Hyperlegible solves exactly that pair: the zero is slashed and the l carries a tail, so Il1 and 0O stop being guesses. It's also the narrowest of the four at 563px.
One number that surprised me. Measured at a 100px em, the x-height lands at 54px for that Arial clone, 53 for Lexend, 51 for Atkinson and 56 for OpenDyslexic, so the "generous x-height" line in most write-ups, mine included until I checked, isn't what separates these fonts. The glyph disambiguation is.
Self-host whichever you pick. The Latin subsets serving this page are 14KB for Lexend, 17KB for Atkinson Hyperlegible and 113KB for OpenDyslexic, and that last number is a real argument for making it an opt-in rather than a default.
Does Line Length Matter More Than Font Choice?
Yes, and this is the part most accessibility checklists skip entirely. Keep body text between 45 and 75 characters per line. Past that, dyslexic readers (and plenty of neurotypical ones, honestly) start losing their place mid-sentence, drifting to the wrong line on the wrap-around. Pair that with line-height at 1.5 to 1.6, tighter than that and lines visually crowd together, looser and the eye has to travel too far to find the next line.
.article-body {
max-width: 65ch;
line-height: 1.6;
letter-spacing: 0.01em;
word-spacing: 0.05em;
}That small letter-spacing and word-spacing bump looks negligible in a code diff. It isn't negligible when you're reading it. Wider spacing gives each letterform room to be recognized individually instead of blurring into its neighbors, which is exactly the cognitive friction dyslexic readers report most often.
There isn't one dyslexia font. There's a set of layout habits that help almost everyone, and a font is just one variable in that set.
What About Color, Contrast, and Background Choices?
Pure white backgrounds behind pure black text create a harsh contrast that some dyslexic readers find genuinely uncomfortable to track for long passages, even though it technically passes WCAG at a strong ratio. A common fix is dropping the background to a warm off-white, something like #FAF7F2, and the text to a near-black rather than pure #000000. You keep a passing contrast ratio while reducing the glare. This is exactly the kind of relationship that belongs in your token system rather than a one-off style override, and our accessibility tokens guide walks through baking that pairing into the design system itself instead of re-deciding it every time a new page ships.
How Do You Audit an Existing Site for This?
Run your body text through the contrast checker on this page first, that's the floor and it's non-negotiable. Then measure your actual rendered line length at a typical viewport width, not your design mockup's assumed width, because fluid layouts often produce longer lines than designers expect on wide monitors. If you're seeing more than 75 characters per line on desktop, that's a bigger accessibility problem than most font debates put together. For the deeper mechanics of measure and rhythm, our line length and vertical rhythm guide is the natural next stop, and it pairs well with the audit steps in our UX audit checklist if you're reviewing a whole product rather than a single article. And if your product ships a dark theme, recheck every text pair there too, because dark-mode typography fails contrast in ways a light-mode audit never surfaces.
Accessible typography for dyslexia isn't a font swap you do once and forget. It's a handful of layout habits, line length, spacing, contrast, applied consistently, that happen to help nearly everyone who reads your site, not just the readers you were specifically trying to accommodate.
Check a pair before you argue about it
Contrast is the one accessibility claim you can settle in four seconds, and it is still the one most often settled by opinion. Worth knowing where it stops helping, though: pure black on pure white clears every threshold here at 21:1 and is reported as uncomfortable by many dyslexic readers. Passing is the floor, not the goal.
