Skip to main content

How to Make a Hebrew or Arabic Website Accessible

The short answer: Hebrew and Arabic website accessibility starts from the same WCAG 2.x AA criteria as any English site. The difference is where things break: language and direction in the code, mixed Hebrew or Arabic and English text, which parts of the layout flip, and how a screen reader with a Hebrew or Arabic voice actually reads the page. An English-only audit rarely looks at any of that, so a site can pass the scan and still be hard to use.

We do this work in Hebrew and Arabic, not as an add-on to English. Over 1,000 projects, for more than 200 authorities, corporations and large businesses. So when an international team tells us the Hebrew version is just the English site with dir="rtl", we have a fair idea of what we’ll find. Here’s what we check, more or less in the order we check it.

Language and direction are two different settings

lang tells the browser and the screen reader which language the text is in. dir tells the browser which way to lay it out. You need both. Set lang="he" without dir="rtl" and you get a Hebrew voice reading a page that runs left to right. Set dir="rtl" without lang and the screen reader may read Hebrew text with an English voice. That comes out as noise, or silence.

Put both on the html element of every page: <html lang="he" dir="rtl"> or <html lang="ar" dir="rtl">. Not on a wrapper div, and not in CSS. W3C is clear on this one: direction belongs in the markup, so it still works when styles don’t load or get overridden.

Then go section by section. An English quote or product description inside a Hebrew page gets its own lang="en". WCAG calls this Language of Parts (3.1.2, level AA). The page language is 3.1.1, level A. Proper names are excluded, so “Google” in a Hebrew sentence is fine.

If the site has both Hebrew and Arabic, each version gets its own lang and dir on its own pages. Write the options in the language switcher in their own language, and mark them: עברית with lang="he", العربية with lang="ar". Otherwise the screen reader announces the Arabic option with the Hebrew voice.

Mixed text: numbers, brand names, phone numbers

Browsers order mixed text with the Unicode bidirectional algorithm. Usually it works. It goes wrong around neutral characters like spaces, plus signs, slashes and periods. Things we keep seeing:

  • A phone number written as +972 3 123 4567 inside a Hebrew or Arabic paragraph. The groups of digits can come out in reverse order, with the plus on the wrong end.
  • “iPhone 15” in a Hebrew sentence that displays as “15 iPhone”.
  • A Hebrew sentence inside a widget that never got dir="rtl", so the period sits on the wrong side and the line reads as broken.

The fix is in the markup. Wrap left-to-right runs in <span dir="ltr">, or <bdi> when the text comes from a database and its direction is unknown. Use dir="auto" on user-generated content.

One thing not to do: don’t type the number backwards in the CMS so it looks right on screen. It will look right, and the screen reader will read it backwards.

What to mirror and what to leave alone

Mirror the things that express direction: the overall layout, back and next arrows, breadcrumb separators, step indicators, the side a drawer menu opens from.

Leave alone the things that don’t: logos, photos (flip a photo and any text in it turns into mirror writing), media play and fast-forward buttons, clocks, and the content of phone numbers, emails, URLs and codes.

Build the layout with CSS logical properties, like margin-inline-start and text-align: start, so it flips with dir by itself. A separate RTL stylesheet full of left and right overrides is where half the bugs come from.

In RTL the “Previous” arrow points right. Its accessible name should say what it does, not where it points.

Focus order and reading order

Tab order follows the DOM. What the user sees follows the CSS. With dir="rtl" on the page, flexbox and grid already run right to left, and the two stay in sync.

The trouble starts when someone takes an LTR template and flips it with flex-direction: row-reverse, order or float: right. On screen it looks correct. Press Tab and focus moves left to right, against everything the user sees. MDN warns about exactly this. It fails WCAG 2.4.3 Focus Order and can fail 1.3.2 Meaningful Sequence, both level A.

In custom tabs, menus and sliders, check that Left and Right arrow keys move the way the user sees it.

Carousels and sliders

Most slider libraries have an RTL option, and dir on the page doesn’t always switch it on. Without it, slides move the wrong way or “next” shows the previous slide.

The buttons need names in the page language. Many libraries ship default screen reader messages like “Next slide” in English, and on a Hebrew site that’s what users hear. Anything that plays by itself needs a pause button (WCAG 2.2.2, level A).

Screen readers with Hebrew and Arabic voices

We test with NVDA and JAWS on Windows, VoiceOver on Mac and iPhone, and TalkBack on Android. NVDA and JAWS can switch voice by the lang attribute, if a voice for that language is installed and the setting is on.

What to listen for: Hebrew or Arabic read with an English voice. English brand names mangled by the Hebrew voice. Prices, dates, phone numbers and abbreviations. And alt text, form labels and ARIA labels that were never translated. On localized sites that’s very common.

Test Hebrew and Arabic separately. Arabic letters connect and change shape by position, Hebrew letters don’t, and Arabic sites may use Eastern Arabic digits. Different fonts, different voices, different bugs.

Fonts, line height and Arabic diacritics

Arabic is taller than Latin text, and its diacritics sit above and below the letters. A line height copied from the English design often clips them, and lines overlap when a user increases spacing. WCAG 1.4.12 Text Spacing (AA) says content must survive a line height of 1.5, paragraph spacing of twice the font size, letter spacing of 0.12em and word spacing of 0.16em. Test it.

Don’t add letter-spacing to Arabic text. It pulls apart letters that are meant to connect. And check that your web font actually includes Arabic glyphs, or the browser falls back to a system font mid-sentence.

Forms

Email, URL, phone, ID and card number fields should be dir="ltr", on Hebrew and Arabic forms alike. Free-text fields get dir="auto", so a user typing English in a Hebrew form sees it in the right order.

Error messages should be in the page language and tied to the field with aria-describedby. The browser’s built-in validation bubbles appear in the browser’s language, so a Hebrew form on an English browser shows English errors. Write the expected date format in the help text, and don’t rely on an asterisk alone to mark a required field.

The standard and the accessibility statement

If the site serves the public in Israel, the requirement is level AA under the Israeli Standard 5568, which is based on WCAG. If you also serve other markets, build and test to WCAG 2.2 AA and cover both in one round of work.

The accessibility statement should be in the language of the site. If you run an Arabic version, it needs the statement in Arabic too. Someone reading the Arabic site shouldn’t have to switch to Hebrew to find out how to report a problem.

What English-first tools and audits usually miss

Recently a nonprofit came to me. It works in Israel and abroad, and its audience speaks Hebrew and Arabic. The site had an automatic accessibility overlay, which gives partial solutions at best. Then a regulatory demand came in to make the site accessible, with a detailed report showing it wasn’t. They were very surprised. They went back to the overlay company and found out the tool wasn’t built at all for sites that aren’t written left to right, like English, French or Italian. That forced them into a full accessibility process at the infrastructure level, the way it has to be done in Hebrew and Arabic. It’s not a simple job. It takes a lot of focused testing, adjusting the navigation order to run right to left, going over every element in every component, and listening to all of it with a screen reader. Sadly, many companies don’t understand what it means to publish a site in Hebrew or Arabic. They look for quick and easy solutions that don’t really fit, and they don’t find out until a complaint or a regulator’s demand lands. The site is built for English, but the content is in Hebrew or Arabic.

English-first audits miss the same things, just more quietly. The auditor checks that lang exists, not that it’s right on every page. They can’t see a scrambled phone number, they test with an English voice, and they don’t notice error messages still in English. A layout flipped with row-reverse looks perfect to them until someone presses Tab.

None of this is exotic. It’s just work that needs someone who reads the language, and that’s the first thing to ask about when you’re choosing a website accessibility consultant.

How we test a Hebrew or Arabic site

We map the unique templates and main user journeys, then test each language version on its own: keyboard, then NVDA, JAWS and VoiceOver with Hebrew and Arabic voices, then mobile. Every issue goes into a report with screenshots and a clear fix. We support your developers, re-test, and write a statement that matches the site.

If you have a website in Hebrew or Arabic and you want to check that it’s accessible, that it meets the regulatory requirements, and that it’s open to every user, with or without a disability, we invite you to reach out for an initial consultation. A high level of accessibility can also help your site get found by search engines and AI tools. We’ll go over your site together, work out which regulations apply, and explain how to meet them in full and how to avoid complaints or lawsuits from users who say the site isn’t accessible. Start with our website accessibility service.

FAQ

No. It’s the starting point. You also need the right lang on every page and section, mixed text marked so it displays in the right order, a focus order that matches the layout, translated labels and messages, and a manual test with Hebrew or Arabic screen reader voices.

Not automatically. Arabic has connected letters, diacritics, different fonts and sometimes different digits, and it’s read with a different voice. Each language version needs its own test, even when both run on the same templates.

They can check a lot of the code. But they won’t spot scrambled numbers or untranslated messages in text they can’t read, or judge how the Hebrew or Arabic voice sounds. Someone on the audit has to work in the language.