The dyslexia font is the one part of that request the research does not support

# The dyslexia font is the one part of that request the research does not support

The request reached you as one line in a ticket: make the site dyslexia-friendly, starting with the font. Whoever filed it was right to file it — somebody on your staff, or somebody who uses your services, is finding your pages hard work. The font is the one item on that list with no measured effect behind it. Size, line height, line length, spacing, and leaving the reader's own settings alone are the items that move reading, they cost nothing to buy, and several of them are already set badly in your stylesheet.

Fifteen studies pooled, 688 readers, and an effect indistinguishable from zero

Annals of Dyslexia published a meta-analysis of dyslexia-friendly typefaces on July 31 2026 — Azzarello, Paek, Hodge, Jameson and Lewis. It pooled 15 studies, 91 effect sizes and 688 readers and asked whether OpenDyslexic, Dyslexie and their relatives improve reading over ordinary faces such as Arial and Times New Roman. The pooled result was g = −0.04, 95% CI [−0.15, 0.07], p = 0.5, or in the authors' own words, "no consistent or reliable effect on reading performance in terms of speed or accuracy" (Azzarello et al., (opens in a new window)Annals of Dyslexia (opens in a new window), online ahead of print, 10.1007/s11881-026-00389-8 (opens in a new window)). That interval straddles zero, so the finding is an absence rather than a reversal, and the paper has not been assigned to an issue yet — worth knowing before you quote it at anybody.

Getting the order of work wrong here has a shape you will recognize. The font ships, the ticket closes, and eight months later a procurement questionnaire asks which WCAG 2.0 AA criteria your site meets. A typeface answers none of them, and the line of 13-pixel grey type in your sidebar — the thing that actually defeated the person who complained — is still there.

Two of the pooled studies answer the objection you will hear first, that the tests were rigged. Wery and Diliberto set their three faces at different nominal sizes on purpose — "Arial 12, OD 10, TNR 12 … varied in order to keep the physical size of the font consistent across probes" — and still found "no improvement in reading rate or accuracy" in twelve diagnosed students aged nine to twelve, none of whom preferred it (Annals of Dyslexia (opens in a new window) 67(2):114–127, 2017 (opens in a new window)). Kuster and colleagues controlled spacing and size across 170 Dutch children with dyslexia, then 102 with and 45 without: they "did not read text written in Dyslexie font faster or more accurately than in Arial font. The majority preferred reading in Arial" (Annals of Dyslexia (opens in a new window) 68(1):25–42, 2018 (opens in a new window)).

[GATE A PLACEHOLDER 1 of 3 — NOT FOR PUBLICATION. Needs the occasion from Mike: the last time somebody asked him to put a dyslexia typeface on a site. Who asked, in what words, what he did first, and what he said back. Interview question 1 in the plan file.]

What moved reading in these studies was the spacing, not the letter shapes

The specialist faces change two things at once: they redraw the letters so similar shapes are harder to confuse, and they push letters and words further apart. Separate the two and the second one carries the effect — which is why the spacing already in your stylesheet is the part worth your attention.

Zorzi and colleagues altered nothing but letter spacing, in an ordinary typeface, with a large unselected sample of Italian and French children with dyslexia, and reported that it "substantially improved text reading performance on the fly (without any training)." Their explanation is crowding: readers with dyslexia are abnormally affected by it, and spacing is what modulates it (PNAS (opens in a new window) 109(28):11455–9, 2012 (opens in a new window)).

Duranovic and colleagues put both variables in one experiment. Dyslexie was no better than Times New Roman once Times New Roman was interspaced, and — the part nobody quotes — "the negative effect of the unfriendly fonts Times New Roman Italic and Curlz MT was eliminated through increased interletter spacing" (Annals of Dyslexia (opens in a new window) 68(3):218–228, 2018 (opens in a new window)). Spacing rescued a face nobody would defend. Kuster's discussion reports the same pattern from Marinus and colleagues in 2016: Dyslexie looked about 7% faster than Arial until the two were matched for within-word and between-word spacing, at which point the advantage disappeared.

The honest version of the vendor claim is that these typefaces ship a spacing change wrapped in a letterform change, and charge you for the letterforms.

Adding letter spacing on its own can make your pages slower to read

Galliussi and colleagues crossed three variables — letterform with or without dyslexia-friendly graphic features, standard or increased inter-letter spacing, standard or increased inter-word spacing — across two groups of 64 Italian children, with and without dyslexia, reading eight equivalent texts aloud. "The data collected failed to show any effect from the letterform," and "reading speed is impaired by an increase in inter-letter spacing not combined with an adequate increase in inter-word spacing" (Annals of Dyslexia (opens in a new window) 70(1):141–152, 2020 (opens in a new window)).

Loosen your tracking and leave word spacing where it was, and the words run into one another. If you touch letter-spacing on body copy, touch word-spacing in the same commit. I have not found that caveat on any of the pages currently ranking for this question, and four of them hand out spacing numbers.

Ontario's requirement names WCAG 2.0 AA, which never names a typeface

Ontario requires websites and web content to conform to WCAG 2.0 Level AA, excluding success criteria 1.2.4 and 1.2.5, for designated public sector organizations and for businesses and non-profits with 50 or more employees, since January 1 2021 (ontario.ca — how to make websites accessible (opens in a new window)). No typeface appears in it, and none appears in WCAG.

What WCAG does specify is close to what the research points at, and two of those criteria are usually reported to you wrongly:

• 1.4.12 Text Spacing carries the numbers everyone quotes — line height 1.5, paragraph spacing 2, letter spacing 0.12, word spacing 0.16, all as multiples of the font size. It is not a design instruction. It requires that "no loss of content or functionality occurs" when the reader applies those values to your page (Understanding 1.4.12 (opens in a new window)): a robustness test of your layout, not a target for your stylesheet. It is also new in WCAG 2.1 (opens in a new window), so Ontario does not require it of you.
• 1.4.8 Visual Presentation holds line length, justification and reader-chosen colour: colours selectable by the user, width no more than 80 characters, "text is not justified," line spacing at least space-and-a-half, resize to 200 percent (Understanding 1.4.8 (opens in a new window)). That one is Level AAA, so Ontario does not require it either.

Which leaves you somewhere uncomfortable and useful: most of the measures with real evidence behind them are ones the law does not oblige you to make. One it does oblige is 1.4.4 Resize Text at Level AA — text resizable to 200 percent without loss of content or functionality (Understanding 1.4.4 (opens in a new window)) — and it is absent from the 2.1 additions, so it is inside what you already owe. If your body copy is pinned in pixels and the layout breaks at 200 percent, one line of CSS is producing a compliance defect and a reading defect at once.

[GATE A PLACEHOLDER 2 of 3 — NOT FOR PUBLICATION. Needs the counted, template-level finding: what Mike has actually found in a stylesheet when he went looking at the type instead of the typeface — how many templates, which value, what one fix closed. Method only if it is employer work, no institutional figures. Interview question 2 in the plan file.]

The reader who needs a different font already has one, unless your CSS overrides it

Somebody who reads more comfortably in OpenDyslexic has usually installed it already, through a browser extension or a reading tool, before ever arriving at your site. Your job in that case is to not fight them.

Three things defeat that reader, and all three are your code rather than their software: a font-family forced with !important on body copy; body text in fixed pixels below the browser default, which ignores the size they set; and a layout that collapses when text grows. Each is an ordinary consequence of a design handed over as a pixel-perfect comp, and what a page does under a reader's own settings takes about ten minutes in a browser to find out. This site publishes its own result, including what has not been tested yet.

Kuster's finding cuts the same way from the other side. When most children with dyslexia in a 170-child sample preferred Arial, a site that forces Dyslexie on everybody has taken a choice away from the people it meant to serve, on the strength of an effect that measured −0.04.

[GATE A PLACEHOLDER 3 of 3 — NOT FOR PUBLICATION. Needs the decision and its cost: something Mike chose or refused on a real project about type or reader control, what it cost, and what happened afterwards that he did not expect. Interview question 3 in the plan file.]

Where the people recommending these fonts are right

Typography is not neutral, and you should not carry the opposite myth home from this. Zorzi measured a change from spacing alone; Duranovic measured a penalty from italics and Curlz MT that spacing then removed. Anyone telling you the question is settled in either direction has read one side of it.

Preference is a real outcome and none of these studies measured it well. They measured rate and accuracy, mostly in children, mostly reading aloud, on word lists and short passages. Not one measured an adult reading your service page for twenty minutes and deciding whether to keep going.

So, two refusals. I will not tell a person who reads better in OpenDyslexic that the evidence says otherwise, because the evidence measured a group average and they are not one. And I will not sell you an accessibility toolbar to close the ticket: the overlay products bundling a font switcher have not delivered compliance and have attracted lawsuits, and a font menu bolted over a 13-pixel stylesheet leaves the stylesheet at 13 pixels.

What to change before you change the typeface

Six items, in the order I would do them, every one cheaper than a font licence:

1. Set body copy at the browser default in relative units. If it is 14px, that is your first defect and probably your largest.
2. Put line height at 1.5 or more on running text.
3. Cap the measure at roughly 80 characters, which is the number 1.4.8 uses.
4. Set text left-aligned, not justified.
5. If you increase letter-spacing, increase word-spacing with it, on Galliussi's finding.
6. Remove anything overriding the reader's own font or size, then confirm the page survives at 200 percent.

If you still want a shortlist of faces, any of the ordinary sans-serifs already on the machine will do, and I am declining to rank them, because the evidence that would justify a ranking does not exist. Arial and Times New Roman are the controls in every study above, which makes them the only two faces the literature can speak to at all, and what it says is that the specialist faces did not beat them. A ranked list of ten would be more useful to you and less true.

If the ticket that started this is really a question about whether your site meets what Ontario requires of it, the font is not where that gets answered. What the law asks of a website, and who it applies to, is a different question with a documented answer — and 13-pixel grey type in a sidebar will fail it long before a typeface does.