Dyslexia-Friendly Fonts in 2026: What the Research Found and Which Ones You May License

Dyslexia-Friendly Fonts in 2026: What the Research Found and Which Ones You May License

A 2026 meta-analysis of 15 studies found no consistent or reliable effect of dyslexia-friendly fonts, such as OpenDyslexic and Dyslexie, on reading speed or accuracy (g = -0.04). OpenDyslexic, Lexend and Atkinson Hyperlegible Next each ship an SIL Open Font License 1.1 file that allows commercial use, with conditions.

The two answers matter to different people. A designer who needs a font for a client project wants the licence terms, and a product team deciding whether to add a "dyslexia mode" wants to know whether the font swap does anything. The pooled interval from the meta-analysis runs from -0.15 to 0.07. Put into reading time with the numbers from one Italian study of children with dyslexia, even the outer edge of that interval is about 14 seconds per 1,000 syllables, against a baseline of roughly 344 seconds.

This is a design and licensing article. It is not a diagnosis, a treatment recommendation or advice about any one child or adult. Decisions about reading support belong with teachers, clinicians and the readers themselves.

What the meta-analysis reports, number by number

The paper is Azzarello, Paek, Hodge, Jameson and Lewis, "Does font improve reading in dyslexic children?: meta-analysis of dyslexia-friendly fonts and dyslexic children's reading performance," published in Annals of Dyslexia (DOI 10.1007/s11881-026-00389-8). Europe PMC lists it as a journal article first published on July 31, 2026, in the 2026 July issue, record PMID 42536336. The article itself sits behind a subscription, so everything below about it is limited to its abstract.

The abstract's central sentence reads: "Using a random-effects model, the results indicate that dyslexia-friendly fonts have no consistent or reliable effect on reading performance in terms of speed or accuracy." Its closing finding adds: "The findings suggest that while certain typographical features may aid legibility, dyslexia-friendly fonts do not offer substantial benefits over traditional fonts for dyslexic readers."

Three details in the table deserve a second look.

Fifteen studies produced 91 effect sizes. That is about six effect sizes per study (91 divided by 15 is 6.07). Meta-analyses commonly take several effect sizes from one experiment, for example one per comparison or outcome. The abstract does not say how the paper handled the fact that effect sizes from one study are not independent of each other, and this article does not guess.

The interval spans zero. A confidence interval of -0.15 to 0.07 is 0.22 units wide. The standard error implied by that width is about 0.056 (0.22 divided by 3.92), and the abstract's z of -0.67 corresponds to a two-sided p of 0.50, which matches the p it reports. The data are compatible with an average difference of exactly zero, and with small differences on either side of it. The largest absolute bound is 0.15.

The abstract does not say which direction the minus sign favours. Authors choose a convention, and the abstract gives none. For that reason this article describes the size of the effect and the width of the interval, and leaves the sign alone. The authors' own word for the effect is "negligible."

What an abstract can and cannot tell you

An abstract fixes the headline and leaves out the machinery. It does not list the 15 studies, so it is not possible to tell from it which of the single experiments described below are in the pool. It does not describe how many of the participants read English and how many read Dutch, Italian, Spanish or another language. It does not report whether the studies were run on paper or on screens. It names speed and accuracy as the outcomes behind the conclusion, and says nothing in those sentences about reading comprehension, even though "Reading Comprehension" appears among the paper's keywords, or about eye strain, stamina or the reader's own preference.

The wording "dyslexic students" in the abstract and "dyslexic children" in the title also leaves the age range open. Treat the result as being about students reading text set in specialist versus common fonts under study conditions. It is not a verdict on every use of every typeface by every reader.

The abstract's last sentence reads: "This meta-analysis provides critical insights for educators and developers of assistive technologies, emphasizing the need for further research to explore alternative strategies for supporting individuals with dyslexia in their academic pursuits." The authors therefore frame the finding as a reason to look elsewhere, not as a reason to stop looking.

Turning g = -0.04 into seconds

A standardised effect size is hard to feel. A rough conversion helps, with the inputs stated so the reader can challenge them. This is an illustration built from outside numbers, not a result from the meta-analysis.

Galliussi and colleagues (2020), an open-access study of Italian children, report that the 64 children with dyslexia read their baseline text at a mean of 2.91 syllables per second with a standard deviation of 0.82. Multiplying each value of g by that standard deviation gives a difference in syllables per second. Dividing 1,000 syllables by the speed gives the reading time, 343.6 seconds at 2.91 syllables per second.

Even the outer bound of the interval is a little over four percent of that group's baseline speed. For scale, the same Italian paper reports that its children with dyslexia read about 35 percent slower than their age-matched peers on the baseline text (2.91 against 4.49 syllables per second, a gap of 1.58). A pooled average effect at the outer interval bound, 0.12 syllables per second, would leave almost all of that gap in place.

The comparison runs the other way too. Galliussi and colleagues judge their own spacing effects to be of no practical importance, since no difference in reading speed between two conditions exceeded 0.20 syllables per second. Against their dyslexia group's mean of 3.18 syllables per second in the test texts, 0.20 is 6.3 percent, or 18.6 seconds per 1,000 syllables (314.5 seconds against 295.9). Differences of that size are the ones the spacing studies below argue over. The outer bound of the meta-analysis interval, converted the same way, is 0.12 syllables per second, under that line.

Two cautions go with this table. A pooled g from 91 effect sizes mixes speed and accuracy outcomes, so the conversion to syllables per second is a convenience and not a claim that every study measured syllables. And the standard deviation comes from one study in one language, so the seconds shift for other readers and other texts.

The single studies behind the headline

The meta-analysis pools experiments, and the experiments are readable on their own. Four of them show what "tested a dyslexia font" has meant in practice.

OpenDyslexic: a small, controlled experiment

Wery and Diliberto used a single-subject alternating treatment design, where each child reads under every condition and the order is randomised. Students read letter lists, real-word lists and nonsense-word lists for one minute each. The probes were printed in Arial at 12, Times New Roman at 12 and OpenDyslexic at 10, because the nominal sizes were adjusted so the physical size of the letters stayed consistent across probes. That detail matters beyond this study: two fonts set at the same nominal size are rarely the same size on the page, and any comparison that skips the adjustment tests size as much as shape.

The authors' abstract concludes: "These results indicate there may be no benefit for translating print materials to this font." Their discussion suggests the font may rest on a common misconception that dyslexia is mainly visual confusion between similar letter shapes, and says four decades of research point to alphabetic and phonological coding difficulties as the probable causes. They also argue that a font change which fails to help can add disappointment for students who have already struggled, and that the time spent converting materials could go to instruction with a better record. Those are the authors' views, offered in a discussion section, and they sit alongside the data without being part of it.

The sample is small, and the authors say that more studies with similar results are needed before anything could be called evidence-based. That is the right reading of a 12-student experiment, and one reason the 2026 pooled analysis is more informative than any single paper.

Dyslexie: two experiments with a lot of children

Kuster and colleagues tested the claim that the Dyslexie font, which was designed specifically for people with dyslexia, eases reading. Their abstract reports three results: no faster or more accurate reading in Experiment 1 (170 children with dyslexia against Arial); no faster or more accurate word-list reading in Experiment 2 (102 children with dyslexia and 45 without, across Dyslexie, Arial and Times New Roman); and a preference for Arial, or for Arial and Times New Roman, over Dyslexie, with preference unrelated to reading performance. Their conclusion in full: "These experiments clearly justify the conclusion that the Dyslexie font neither benefits nor impedes the reading process of children with and without dyslexia."

Their method section is a lesson in controlling a font comparison. Dyslexie sets taller letters with extra built-in line space, so the team set Arial at body size 13 with a vertical line spacing of 1.35 and Dyslexie at body size 12 with a line spacing of 1.0. The reported capital heights were 3.28 millimetres for Arial and 3.95 for Dyslexie, with x-heights of 2.38 and 2.51 millimetres. They did this because the study was meant to isolate letter shape, not size or interline space.

Marinus: where the 7 percent went

Marinus and colleagues went looking for what a positive Dyslexie result could rest on. With 39 low-progress readers, Dyslexie did beat standard Arial, by 7 percent more words per minute. When Arial was given Dyslexie's within-word and between-word spacing, the difference in reading speed was no longer significant. The authors' conclusion: "We concluded that the efficacy of Dyslexie font is not because of its specially designed letter shapes, but because of its particular spacing settings."

Set beside Kuster's result, the same Dyslexie font showed no benefit against Arial for 170 children and a modest benefit against tightly spaced Arial for 39, and the second benefit vanished when spacing was equalised. Both outcomes fit the claim that letter shape adds nothing and spacing might.

Galliussi: separating shape from spacing

Galliussi and colleagues designed their experiment to pull the two apart. Their font was built for the study, close to the Italian EasyReading design, with asymmetric shapes in the b-d-p-q group, mixed serif and sans serif letters and longer ascenders and descenders. They crossed letterform with letter spacing and with word spacing, so each child read eight texts in eight conditions. The abstract states: "The data collected failed to show any effect from the letterform. As regards spacing, the data showed that reading speed is impaired by an increase in inter-letter spacing not combined with an adequate increase in inter-word spacing."

Preference is data too

Two of these studies recorded what children liked. In Wery and Diliberto, some students called OpenDyslexic "new" or "different," and none of the participants reported preferring to read in it. In Kuster's two experiments, preference went to Arial and Times New Roman and was not related to reading performance. An individual reader might still prefer a specialised font, and nothing here says that preference is wrong. What the studies do say is that a site which switches its default body font on the assumption that readers with dyslexia will like it is making a bet the two experiments did not support, and that the better design is a setting the reader controls.

Spacing: the live issue inside the "dyslexia font" idea

Strip the letterforms away and the hypothesis these studies test is crowding. Galliussi's introduction explains the reasoning: some researchers have linked dyslexia to visual crowding, the difficulty of identifying a letter when others surround it, and spacing is the typographic lever for crowding. The results, though, are mixed. The table collects what each abstract reports.

The Galliussi paper reviews the field before reporting its own result, and its review is candid: it calls the evidence for a spacing effect controversial. It describes the Zorzi experiment as enlarging Times New Roman letter spacing by 2.5 points on a 14-point body size, about 18 percent of the body size, and observes that the compared sentences were not matched in length, which matters because short lines are easier to read. It notes that Hakvoort and colleagues, using sentences of the same length, found fewer errors with extra-large spacing but no gain in speed. And it offers possible reasons its results differ from earlier ones. Against Zorzi's, the candidates are the line-length confound and the larger amount of added spacing. Against Bachmann's, the candidate is age: its children were in the first year of Italian secondary school and Bachmann's were in the fourth year of elementary school, so one hypothesis is that wide spacing helps while reading is being acquired and matters less afterwards. These are hypotheses in the paper, not findings.

The Galliussi result most useful to anyone setting type is the one about the mismatch. Their increased-spacing condition added 0.07 em to letter gaps and 0.27 em to word gaps. The condition that slowed readers was the one with the letter gaps widened and the word gaps left alone. The authors' explanation is that when the space inside words grows toward the size of the space between words, the words are harder to separate. For readers with dyslexia the slowdown was a drop from 3.22 to 3.04 syllables per second, and the typical readers slowed as well. The authors call the effect small, and 0.18 syllables per second sits inside the 0.20 they treat as not practically meaningful, so the mismatch is a hint about direction, not a measured threshold.

Using the paper's own typographic ranges, a normal letter gap runs from 0 to 15 percent of the body size and a normal word gap about 20 to 25 percent. The paper sets its default spacing from Verdana Pro Condensed and gives no gap values for it, so this is a model: if the general ranges described that font, the smallest word gap (0.20 em) sat against a largest letter gap of 0.15 plus 0.07, which is 0.22 em, a ratio of 0.91. In ordinary text the same ratio is 0.20 over 0.15, or 1.33. The word gap has stopped being clearly larger than the letter gap.

Stagg and Kiss, and Sjoblom and colleagues, found that spacing helped older readers, adolescents and adults, including measured speed gains. Both tested extra-large letter spacing, and Stagg and Kiss conclude in their abstract that increasing letter spacing is an effective way for teachers to improve reading skills in students with dyslexia. The sensible summary across the five studies is that some experiments find gains from added spacing, others find small or no gains, and the manner of adding it can reverse the result.

Print size and line length

Two other typographic variables have their own support. They are less glamorous than a font name, and any site can set them.

Size. O'Brien, Mansfield and Legge (2005) measured reading speed across twelve print sizes in dyslexic readers in grades 2 to 4 and in younger non-dyslexic readers matched for word reading ability. Both groups' reading curves had the same two-limbed shape, flat across large sizes and falling sharply below a critical print size. The dyslexic curves showed higher critical print sizes. That suggests setting text comfortably above the size where speed starts to fall, though the abstract gives no recommended size. Galliussi's materials used 14-point type on A4 paper, a size the paper attributes to British Dyslexia Association guidelines.

Line length. Schneps and colleagues (2013) compared reading on paper with reading on a small handheld device in 103 high school students with dyslexia. Paper lines averaged 13.94 words (14-point Times); the handheld screen set 42-point Times with an average of 3.40 words per line. Speed and comprehension improved on the device for specific subsets of readers: those who struggled most with phoneme decoding or sight-word reading read faster, and those with limited visual attention spans gained in comprehension. The authors suggest it was the short lines, not the device, that mattered, and report an earlier eye-tracking experiment from their laboratories with 26 students in which reading speed rose 27 percent on lines of 2.19 words against 11.6 words.

Neither result translates to a web column. A 3.4-word line is about a quarter of the paper line (3.40 over 13.94 is 24 percent) and no one would typeset an article that way. What the research supports is a narrower claim: shorter lines helped some readers with dyslexia, and the effect was larger for those with the greatest difficulty. WCAG's Level AAA criterion 1.4.8, Visual Presentation, asks for line width of no more than 80 characters, text that is not justified, line spacing of at least one and a half within paragraphs, and paragraph spacing at least 1.5 times the line spacing. The Understanding document says narrow text blocks make it easier for some readers to continue on to the next line, and that fully justified text can create "rivers of white" space. That is a W3C rationale, not a trial, and the 80-character limit is a ceiling, not a target.

What three licence files say

Licensing is the part of this topic with a definitive answer, because the answer is a text file. The table below is taken from each project's own OFL.txt, the file that ships in its repository.

All three open with the statement "This Font Software is licensed under the SIL Open Font License, Version 1.1." From the licence title line to the disclaimer, the licence text in the three files is identical, so one set of conditions applies to all three. The only differences are the copyright lines and the Reserved Font Name.

Dyslexie is not covered in this table. Its licence text was not available to review, so no licence is asserted for it here. Get the terms from its distributor before using it anywhere.

What the OFL permits, and its conditions

The grant is one sentence: "Permission is hereby granted, free of charge, to any person obtaining a copy of the Font Software, to use, study, copy, merge, embed, modify, redistribute, and sell modified and unmodified copies of the Font Software, subject to the following conditions:" Nothing in it separates commercial from personal use, so commercial use is inside the grant. Five conditions follow.

1. "Neither the Font Software nor any of its individual components, in Original or Modified Versions, may be sold by itself."
2. "Original or Modified Versions of the Font Software may be bundled, redistributed and/or sold with any software, provided that each copy contains the above copyright notice and this license. These can be included either as stand-alone text files, human-readable headers or in the appropriate machine-readable metadata fields within text or binary files as long as those fields can be easily viewed by the user."
3. "No Modified Version of the Font Software may use the Reserved Font Name(s) unless explicit written permission is granted by the corresponding Copyright Holder. This restriction only applies to the primary font name as presented to the users."
4. "The name(s) of the Copyright Holder(s) or the Author(s) of the Font Software shall not be used to promote, endorse or advertise any Modified Version, except to acknowledge the contribution(s) of the Copyright Holder(s) and the Author(s) or with their explicit written permission."
5. "The Font Software, modified or unmodified, in part or in whole, must be distributed entirely under this license, and must not be distributed under any other license. The requirement for fonts to remain under this license does not apply to any document created using the Font Software."

The licence also contains a termination clause, "This license becomes null and void if any of the above conditions are not met," and the standard disclaimer of warranty. The definitions set the scope of condition 3. The licence defines the term this way: Reserved Font Name refers to any names specified as such after the copyright statement(s). That is why the table separates the three files. The Atkinson Hyperlegible Next file specifies none, so condition 3 has nothing to attach to. The OpenDyslexic file reserves "OpenDyslexic." The Lexend file reserves "RevReading Lexend," not the shorter family name Lexend. Whether a modified build may use a name that contains the word Lexend is a question for the copyright holder, and the cautious practice is a new name for any modified build.

What this means for a website

For a site, the licence question usually reduces to files you serve. Three points come from the official OFL FAQ (version 1.1-update7, November 2023, on openfontlicense.org).

• Using the fonts on web pages is allowed, and the FAQ calls loading them through @font-face recommended and explicitly allowed by the licensing model, because it counts as distribution. Distribution triggers condition 2, so the licence text and copyright notice travel with the font files you host.
• A change of format is normally considered modification. A WOFF or WOFF2 file escapes that only when the original font data is unchanged apart from compression and any WOFF metadata is either omitted or carries over the original metadata unaltered. The FAQ warns that some conversion tools do not meet that bar, and their output must be treated as a Modified Version.
• Subsetting is modification. FAQ 2.6 answers: "Yes. Removing any parts of the font when delivering a webfont to a browser, including unused glyphs and smart font code, is considered modification. This is permitted by the OFL but would not normally allow the use of RFNs. Some newer subsetting technologies may be able to subset in a way that allows users to effectively have access to the complete font, including smart font behaviour." The FAQ points to its entry 2.8 for that last case.

Put together, a subsetted WOFF2 of OpenDyslexic or Lexend that you built yourself should carry a new name, and you keep the copyright notice and the licence text with it. A subsetted Atkinson Hyperlegible Next has no reserved name to drop, though the copyright and licence text still travel with it and the "not sold by itself" rule still applies. None of the three licences touches the documents or pages you set in the font: condition 5 says the requirement does not apply to a document created with the Font Software, and FAQ 1.1 answers yes to using an OFL font for logos, posters and similar design work, stating that no additional licence or permission is required. Our guides to commercial font licensing and the design asset licensing checklist cover the surrounding questions for client work, and the guide to self-hosting Google Fonts walks through the hosting side.

A few cautions about the files themselves. OpenDyslexic's site states: "The newest version of OpenDyslexic now uses the SIL-OFL license, giving you freedom to use it for personal use, business use, education, commercial, books, ebook readers, applications, websites, and any other project or purpose you need." The word "newest" implies earlier releases were distributed on other terms. If a copy came from a bundle, a theme or a download site, open the licence file inside that download. The Atkinson file cited here belongs to the Atkinson Hyperlegible Next repository, so check the licence file that comes with any other Atkinson download. Also, the OpenDyslexic sources have moved from GitHub to a Forgejo host (forge.hackers.town), and the OFL.txt there is identical to the GitHub copy. A licence is a legal text, not legal advice, and a contract with a client may add its own requirements.

What each project says about its own purpose

The three projects describe their goals differently, and the differences are useful when you write documentation or marketing.

Atkinson Hyperlegible Next: its README says: "Atkinson Hyperlegible, named after the founder of the Braille Institute, has been developed specifically to increase legibility for readers with low vision, and to improve comprehension." It names low vision as the target group, and the README does not mention dyslexia anywhere. The README records the first release of this version, 2.001, on November 20, 2024, and describes the family as an improvement on the original typeface with new characters, improved glyphs and kerning, and six weights in upright and italic.

OpenDyslexic: the project's site says "OpenDyslexic is a typeface designed against some common symptoms of dyslexia." That is the design claim the research above tested, in the form of the thick-bottom letterforms Wery and Diliberto described.

Lexend: the README states: "Applying the Shaver-Troup Individually Optimal Text Formation Factors, studies have found readers instantaneously improve their reading fluency." It also says the font is based on Quicksand and that "Thomas modified Quicksand for the specialized task of improving reading fluency in low-proficiency readers (including those with dyslexia.)" The sentence about studies carries no citation in the README, so it is the project's statement, not something the meta-analysis abstract covers.

On the research side, the meta-analysis abstract names OpenDyslexic and Dyslexie as its examples. A Europe PMC search on November 3, 2026 for Lexend together with font or typeface and reading returned zero records, and a search for the phrase "Atkinson Hyperlegible" returned zero as well. That search covers one index and says nothing about studies published elsewhere. It does mean that index offers no paper to quote for or against either font as a dyslexia aid, so the evidence-based description of both is openly licensed sans serif families with stated design goals.

CSS settings for body text, with the arithmetic

Where the research points, even weakly, is toward adjustable spacing and a comfortable size, not a particular font name. The settings below separate three kinds of number: values from the WCAG text spacing test, values from the experiments, and values that are simply good defaults. All arithmetic assumes a 16-pixel root font size and, as a second column, the 18.67-pixel size that results from converting 14 points.

A caution on the first row. Fourteen points on A4 paper at reading distance is not the same visual size as 18.67 CSS pixels on a phone or a monitor, because viewing distance and pixel density differ. The conversion is arithmetic, and the number is a reference point, not an equivalence.

How CSS letter spacing treats the gaps between words

The CSS Text Module Level 3 (Candidate Recommendation Draft, August 14, 2026) defines letter-spacing this way: "This property specifies additional spacing (commonly called tracking) between adjacent typographic character units." It adds: "Letter-spacing is applied to each typographic character unit by inserting half of the additional spacing on each side; except it is not applied at the beginning or end of a line." A space is a character unit and also a word separator. The word-spacing property adds its length to each word separator, applied half on each side of the character unless typographic tradition dictates otherwise. Under that model, a letter-spacing value L widens every letter gap by L and every word gap by 2L (half before the space, half after, plus the neighbouring letters' halves), and word-spacing W adds W on top of that.

Using Galliussi's typographic ranges, with letter gaps up to 0.15 em and word gaps of 0.20 to 0.25 em, the model gives these gaps. The ratio is the smallest word gap divided by the largest letter gap, so higher means words stay clearly separated.

This table is a model, built from the CSS Text Level 3 wording and Galliussi's ranges, and individual fonts and browsers differ. Its use is directional. Widening tracking with CSS letter-spacing also widens the space characters, so the 0.91 situation of the slow condition is hard to reach by accident, and adding a word-spacing value on top keeps a wide margin. Check it by measuring a sample paragraph in your own font before relying on it.

The CSS Text draft adds that when letter-spacing is non-zero, user agents should not apply optional ligatures, which is why its own example shows the fi in "filial" staying unjoined; features set explicitly through font-feature-settings take precedence. Designers who rely on fi and fl ligatures in body text will see them drop out whenever tracking is on.

A reader-controlled setting

The text spacing criterion is a statement about what must not break, and it shows how to build the control. WCAG 1.4.12 requires that no content or functionality is lost when a user sets line height to at least 1.5 times the font size, paragraph spacing to at least 2 times, letter spacing to at least 0.12 times and word spacing to at least 0.16 times. Its Note 1 says content is not required to use these values; the requirement is that nothing is lost when a user overrides the authored spacing. Its Benefits section says: "People with dyslexia may increase space between lines, words, and letters to increase reading speed."

A setting that ships those values behind a toggle gives readers the lever the studies tested. The toggle should be a user preference, saved on the device, and not a hidden default.

:root {
  --read-size: 1.1667rem;     /* 18.67px at a 16px root */
  --read-line: 1.5;           /* 28px at 18.67px */
  --read-letter: 0.12em;      /* 2.24px at 18.67px */
  --read-word: 0.16em;        /* 2.99px at 18.67px */
  --read-para: 2em;           /* 37.33px at 18.67px */
}
[data-reading="spaced"] .prose {
  font-size: var(--read-size);
  line-height: var(--read-line);
  letter-spacing: var(--read-letter);
  word-spacing: var(--read-word);
  max-width: 80ch;
  text-align: left;
}
[data-reading="spaced"] .prose p { margin-block: 0 var(--read-para); }

To test that the layout survives, apply the four WCAG values with !important to every element on a copy of the page and look for clipped text, overlapping lines and controls that moved out of reach. Fixed heights on buttons, cards and menu bars are the usual cause of failures. Our px, rem and em guide explains why em-based spacing follows the text size, and the typography scale guide shows where a reading size sits in a type system. The WCAG 2.2 compliance guide places this criterion among the others a designer owns.

A decision rule for a site

The pieces fit into a short sequence, each step tied to something above.

1. Pick the body font on its own merits. The meta-analysis gives no basis for choosing a font because it is marketed for dyslexia. Atkinson Hyperlegible Next, Lexend and a plain system sans are all workable body fonts. Choose among them for rendering quality, language coverage, file size and licence terms.
2. Do not tell readers a font will help them. The pooled result and the four single experiments above show no benefit from letterforms themselves, and the one positive Dyslexie finding disappeared when spacing was matched, so none of them supports a claim that a font improves reading for people with dyslexia. If your page offers OpenDyslexic as a choice, say it is a choice, and do not describe an effect.
3. Build the spacing, size and width controls. This is the part with some support in the studies and direct support in WCAG 1.4.12 as a user need. Give the controls a visible place, remember the setting, and apply letter and word spacing together.
4. Make the layout survive user overrides. Test with the four WCAG 1.4.12 values. Many readers will use a browser extension or a user stylesheet instead of your toggle, and a layout that clips text defeats both.
5. If you offer a specialised font, license it correctly. Self-host the files, ship the OFL.txt, keep the copyright notice with the font, rename any modified build when the file reserves a name, and keep an inventory entry for each font naming the licence file you read and when.
6. Ask people who read your content. In Kuster's experiments, preference was not related to reading performance. A short survey or a few moderated sessions with readers will tell you more about your own pages than any pooled average.

What no study here settles

The meta-analysis tests fonts as a class against Arial and Times New Roman, in experimental reading tasks, measured by speed and accuracy. It cannot tell you about long-term reading stamina, about choice and comfort in daily life, or about any reader's own best setup. Individual readers may differ, and a reader may prefer one font over another. Taking a pooled result seriously and respecting an individual's choice are compatible. The first is a reason not to make a claim for a font on behalf of everyone, and the second is a reason to let each person pick.

Dyslexia font questions

Is OpenDyslexic proven to help people with dyslexia read faster?

No. The 2026 meta-analysis of 15 studies reports no consistent or reliable effect of dyslexia-friendly fonts such as OpenDyslexic and Dyslexie on reading speed or accuracy, with an overall g of -0.04 and a 95% confidence interval from -0.15 to 0.07. A separate 12-student experiment on OpenDyslexic found no improvement in rate or accuracy and no student who preferred the font.

Does the meta-analysis mean nothing helps readers with dyslexia?

No. The paper tests one thing, specialised fonts against common fonts. Its abstract says certain typographical features may aid legibility and calls for further research on other strategies. Studies of letter spacing report gains in some groups, and the evidence on spacing is mixed. Decisions about support for a particular reader are for that reader, their teachers and specialists.

Can I use OpenDyslexic, Lexend or Atkinson Hyperlegible Next in a commercial project?

The OFL.txt files for all three grant permission to use, study, copy, merge, embed, modify, redistribute and sell copies of the font, free of charge, subject to five conditions: the font cannot be sold by itself, the copyright notice and licence travel with copies, modified versions cannot use a Reserved Font Name without permission, the authors' names cannot promote a modified version, and the font stays under the OFL. Check the licence file inside the specific download you use.

Do I have to rename a font if I convert it to WOFF2?

Only if the conversion counts as a modification and the font reserves a name. The OFL FAQ says a WOFF or WOFF2 file is not a modification when the font data is unchanged apart from compression and the metadata is carried over intact, and that some conversion tools do not meet that standard. OpenDyslexic reserves "OpenDyslexic" and Lexend reserves "RevReading Lexend." The Atkinson Hyperlegible Next file reserves no name.

Is Atkinson Hyperlegible a dyslexia font?

Its README describes the typeface as developed specifically to increase legibility for readers with low vision and to improve comprehension. The README does not mention dyslexia. The meta-analysis abstract names OpenDyslexic and Dyslexie as its examples and does not mention Atkinson Hyperlegible.

What spacing settings did the studies use?

Zorzi and colleagues enlarged Times New Roman letter spacing by 2.5 points on a 14-point body, about 18 percent of the body size, or 0.1786 em. Galliussi and colleagues added 0.07 em to letter gaps and 0.27 em to word gaps and found that widening letters without widening words slowed reading. WCAG 1.4.12 uses 0.12 em letter spacing and 0.16 em word spacing as the values a layout must survive.

Studies and licence files cited

• Azzarello C, Paek S, Hodge C, Jameson M, Lewis J. Does font improve reading in dyslexic children?: meta-analysis of dyslexia-friendly fonts and dyslexic children's reading performance. Annals of Dyslexia, 2026 Jul; DOI 10.1007/s11881-026-00389-8. Abstract as indexed by Europe PMC (PMID 42536336); checked November 3, 2026. The full article requires a subscription, and this page draws only on the abstract.
• Wery JJ, Diliberto JA. The effect of a specialized dyslexia font, OpenDyslexic, on reading rate and accuracy. Annals of Dyslexia 2017;67(2):114-127; DOI 10.1007/s11881-016-0127-1. Open-access full text (CC BY 4.0) via Europe PMC, PMC5629233; checked November 3, 2026.
• Kuster SM, van Weerdenburg M, Gompel M, Bosman AMT. Dyslexie font does not benefit reading in children with or without dyslexia. Annals of Dyslexia 2018;68(1):25-42; DOI 10.1007/s11881-017-0154-6. Open-access full text via Europe PMC, PMC5934461; checked November 3, 2026.
• Marinus E, Mostard M, Segers E, Schubert TM, Madelaine A, Wheldall K. A special font for people with dyslexia: does it work and, if so, why? Dyslexia 2016;22(3):233-244; DOI 10.1002/dys.1527. Abstract via Europe PMC (PMID 27194598); checked November 3, 2026.
• Galliussi J, Perondi L, Chia G, Gerbino W, Bernardis P. Inter-letter spacing, inter-word spacing, and font with dyslexia-friendly features: testing text readability in people with and without dyslexia. Annals of Dyslexia 2020;70(1):141-152; DOI 10.1007/s11881-020-00194-x. Open-access full text (CC BY 4.0) via Europe PMC, PMC7188700; checked November 3, 2026.
• Zorzi M, et al. Extra-large letter spacing improves reading in dyslexia. PNAS 2012;109(28):11455-11459; DOI 10.1073/pnas.1205566109. Abstract via Europe PMC (PMID 22665803); the 2.5-point spacing figure is as described in Galliussi et al.; checked November 3, 2026.
• Sjoblom AM, Eaton E, Stagg SD. The effects of letter spacing and coloured overlays on reading speed and accuracy in adult dyslexia. British Journal of Educational Psychology 2016;86(4):630-639; DOI 10.1111/bjep.12127. Abstract via Europe PMC (PMID 27629067); checked November 3, 2026.
• Stagg SD, Kiss N. Room to read: the effect of extra-large letter spacing and coloured overlays on reading speed and accuracy in adolescents with dyslexia. Research in Developmental Disabilities 2021;119:104065; DOI 10.1016/j.ridd.2021.104065. Abstract via Europe PMC (PMID 34600780); checked November 3, 2026.
• O'Brien BA, Mansfield JS, Legge GE. The effect of print size on reading speed in dyslexia. Journal of Research in Reading 2005;28(3):332-349; DOI 10.1111/j.1467-9817.2005.00273.x. Abstract via Europe PMC (PMID 16601822); checked November 3, 2026.
• Schneps MH, Thomson JM, Chen C, Sonnert G, Pomplun M. E-readers are more effective than paper for some with dyslexia. PLoS ONE 2013;8(9):e75634; DOI 10.1371/journal.pone.0075634. Open-access full text via Europe PMC, PMC3776763; checked November 3, 2026.
• Atkinson Hyperlegible Next project, OFL.txt and README, main branch; OFL.txt last changed November 20, 2024; checked November 3, 2026.
• OpenDyslexic project, OFL.txt (last changed July 29, 2019; identical copy on the project's current host, forge.hackers.town) and the site opendyslexic.org; checked November 3, 2026.
• Lexend project, OFL.txt (last changed July 30, 2021) and README; checked November 3, 2026.
• SIL International, OFL-FAQ version 1.1-update7, November 2023, entries 1.1, 2.1, 2.2 and 2.6; checked November 3, 2026.
• W3C, Understanding SC 1.4.12: Text Spacing, criterion text, Note 1 and Benefits; checked November 3, 2026.
• W3C, Understanding SC 1.4.8: Visual Presentation, Level AAA, updated October 1, 2025; checked November 3, 2026.
• W3C, CSS Text Module Level 3, Candidate Recommendation Draft of August 14, 2026, letter-spacing and word-spacing; checked November 3, 2026.
• Scope note: no licence text for Dyslexie was available, so no licence is stated for it. The Europe PMC searches for Lexend and for Atkinson Hyperlegible cover that one index only.