Chat: Markdown in AI replies not rendering (bold and code show as raw syntax)

Support Bug Reports
1

Where does the bug appear (feature/product)?

Cursor IDE

Describe the Bug

In the AI chat panel, Markdown in the assistant’s replies is not being rendered. I see the raw characters instead of formatted output:
Bold text appears as double asterisks instead of bold.
Inline code that includes a backtick (e.g. the shortcut “Ctrl+backtick”) breaks or the backtick disappears, because the backtick is interpreted as closing the code span instead of being shown as the key name.
So the chat is treating the reply as plain text rather than parsing it as Markdown.

Request ID: 7b826520-6853-4326-a9fa-ddb5875700eb

Steps to Reproduce

  1. enter the following in chat: how to show inline code vs code block in markdown
  2. see how the response rendered incorrectly for me in the attached screenshot

Expected Behavior

Properly rendered markdown. I would attached an example from vs code preview of the cursor output, but it doesn’t look like I can attached more than one image.

Screenshots / Screen Recordings

Screenshot 2026-02-21 at 11.05.33 AM.png
Screenshot 2026-02-21 at 11.05.33 AM.png1674×1230 212 KB

Operating System

MacOS

Version Information

Version: 2.5.20
VSCode Version: 1.105.1
Commit: 511523af765daeb1fa69500ab0df5b6524424610
Date: 2026-02-19T20:41:31.942Z
Build Type: Stable
Release Track: Default
Electron: 39.4.0
Chromium: 142.0.7444.265
Node.js: 22.22.0
V8: 14.2.231.22-electron.0
OS: Darwin arm64 25.3.0

For AI issues: which model did you use?

Auto

For AI issues: add Request ID with privacy disabled

Request ID: 7b826520-6853-4326-a9fa-ddb5875700eb

Does this stop you from using Cursor

No - Cursor works, but with this issue

3

Hey, thanks for the detailed bug report and the screenshot. I can see that raw Markdown characters are showing up instead of rendered formatting.

This looks like an edge case in the chat Markdown parser, especially when the model’s reply includes backticks (for example, Ctrl+backtick). The parser treats them as inline code markers, and that can trigger a chain reaction that breaks the rest of the formatting.

A couple questions:

  1. Does this happen every time, or only with requests where the reply is likely to include backticks?
  2. Have you tried selecting a specific model (for example, Claude 4.6 Sonnet or GPT-5.3) instead of Auto, to compare the formatting?

This is a known area to improve in the Markdown renderer. I’ve shared your report with the team. There’s no ETA yet, but your report helps us prioritize.

Let me know how it goes with a specific model.

5

  1. So far it seems specifically related to backticks. My search of the forums shows that
    others have reported this previously.

  2. The markdown that the models are generating is fine. To confirm this, I used Copy Message (in Cursor chat) and pasted the message (raw markdown) into VS Code; the preview mode renders the markdown correctly.

    But, just to confirm, I tried composer, opus and gpt codex. They all output correct markdown (per testing in VS Code).

    It definitely appears to be an issue with the Cursor app (also broken in the browser/cloud-based Cursor).

7

@johncaruso Let us know if you’re still facing an issue! I have not been able to reproduce recently (2.6)

Screenshot 2026-03-25 at 15.29.31
Screenshot 2026-03-25 at 15.29.311648×692 30 KB

8

I was having the same issue on the latest version at the time of writing. I tried downgrading to 2.6.11 but the issue was still present. Then tried 2.5.26 - markdown rendering issue went away. The regression seems to start sometime after that.

9

@tritonality

If you have some examples of models/prompts used to reproduce, please share them! I have not been able to reproduce on my own.