RTL (Right-to-Left) text is rendered incorrectly as LTR

Open 💬 3 comments Opened Jul 15, 2026 by alighafoorzade

💡 Likely answer: A maintainer (github-actions[bot], contributor) responded on this thread — see the highlighted reply below.

What version of the Codex App are you using (From “About Codex” dialog)?

26.707.72221

What subscription do you have?

Pro

What platform is your computer?

x64

What issue are you seeing?

When using the ChatGPT Desktop application, right-to-left (RTL) languages such as Persian (Farsi) are not rendered correctly.

Instead of displaying the text from right to left, the application lays out the content as left-to-right (LTR), causing punctuation, mixed English/Persian text, code snippets, and paragraph alignment to appear incorrectly.

What steps can reproduce the bug?

Feedback ID: no-active-thread-019f65a5-5d4f-7423-85ee-6d1e63f5d241

What is the expected behavior?

RTL paragraphs should be rendered with proper right-to-left direction.
Mixed RTL/LTR text (e.g., Persian with English words, code, or numbers) should follow the Unicode Bidirectional Algorithm and display correctly.
Text alignment and cursor behavior should match standard RTL text rendering.

Additional information

_No response_

View original on GitHub ↗

3 Comments

github-actions[bot] contributor · 1 month ago

Potential duplicates detected. Please review them and close your issue if it is a duplicate.

#32912
#32490

Powered by Codex Action

rshouker · 1 month ago

Now that this the ChatGPT app, and the old one supported it, I consider it a major regression. Please fix it ASAP, because it makes it unsuable.

anasalrawhani · 29 days ago

Confirmed on Arabic in the current macOS desktop build.

Environment: ChatGPT Desktop / Codex surface 26.721.81911 (build 5973), macOS 26.5.2 (25F84), Apple Silicon arm64.

This affects more than paragraph alignment. When Arabic explanations contain English technical terms such as Deep Learning, Big Data, Data Privacy, and Human Oversight, the surrounding Arabic sentence is visually reordered. Terms, punctuation, and wrapped lines appear beside the wrong parts of the sentence, making the content extremely difficult to read.

I also tested pure Arabic paragraphs, Markdown headings and lists, right-aligned Markdown tables, inline code, Unicode direction marks, directional isolation characters, and explicit RTL HTML containers. None produced correct right-aligned rendering with stable mixed Arabic/English order.

This is a high-impact internationalization and accessibility regression for Arabic-speaking developers because technical conversations necessarily mix Arabic prose with English identifiers, package names, commands, paths, and code. Please escalate this to the desktop renderer and internationalization team. Screenshots from the current build are attached.