Arabic questions need their own visibility reading

Arabic & Persian Layout Requirements is a W3C Group Draft Note from the Internationalization Working Group, dated 2 October 2025. The note says it is a draft, may be replaced, and should not be cited as a finished standard. With that limit stated, the text still records a layout fact: Arabic script is written from right to left; numbers, including Arabic numbers, are written from left to right; and when Arabic script is the main script, the page layout is also right to left. Latin letters are a left-to-right class in the bidirectional algorithm.

An Arabic sentence that includes an English brand name changes direction inside the sentence. That is a layout requirement, not evidence that a model will cite the page.

Localized versions of your pages is about Google Search. hreflang tells Google about language or regional versions of a page. The document says Google does not use hreflang or the HTML lang attribute to detect the language of a page; it uses its own algorithms. A country code cannot stand alone as the first part of an hreflang value. Adding an Arabic URL to an English page does not turn the English body into Arabic. The document does not say ChatGPT or Perplexity will cite the page because of the tag.

inLanguage is the schema.org property for the language of the content. The property page says to use an IETF BCP 47 language code. It is a vocabulary property, not a citation guarantee.

The Common Crawl language table reports monthly crawl archives. CLD2 identifies the language and can mark up to three languages in one document. The table is the percentage for the primary language, and only HTML pages are sent to the detector. In the columns CC-MAIN-2026-30, CC-MAIN-2026-34, and CC-MAIN-2026-39, Arabic (ara) is 0.6548%, 0.6677%, and 0.6679%. English (eng) is 40.5782%, 40.4526%, and 41.8620%. Chinese (zho) is 4.4292%, 4.3829%, and 4.2255%.

A bilingual page whose primary language is not Arabic is not counted in the ara row. These figures describe public HTML in those archives. They are not a citation share for ChatGPT or Perplexity, and they are not the language of questions from people in Saudi Arabia or the UAE. Later archives will differ. Use the current table.

The OpenAI web search guide documents country as a two-letter ISO country code. The Perplexity Sonar schema separates search_language_filter and language_preference as ISO 639-1 fields, and documents user_location.country as ISO 3166-1 alpha-2. Neither document says to merge the UAE and Saudi Arabia, or that an English page stands in for an Arabic question.

Pooling those conditions produces a pooled report. It does not match the fields above. Arabic’s ISO 639-1 code is ar. The two-letter country codes for the UAE and Saudi Arabia are AE and SA. The documents’ examples do not use those three codes. A snapshot should still record the codes that were actually sent.