In custom GPT, some images in Markdown format cannot be displayed irregularly, but they can be reproduced stably under the same conditions

I’ve been using my custom GPT to display images from external networks in Markdown format without any issues for over a month. However, recently, I’ve encountered a problem where these images are not showing up. This issue doesn’t occur in other contexts, where the same images are displayed correctly. The problem is consistent under the same conditions, but it varies across different scenarios, making it hard to pinpoint a specific cause. Here are a few examples:

  1. In my Amazing Girlfriends GPT at , entering /intro should display the Markdown texts of two images:
    ![**Critical Visual - Must Be Displayed**](
    Wechat Group: ![](
    Yet, they don’t appear.

  2. When I prompt this GPT to replicate the aforementioned output, it displays the following image as per my settings (which prevent any content copying and instead show a female ghost image):
    ![Sadako Yamamura](

  3. Interestingly, when I repeat the output from the first GPT in another GPT ( ), even though the text output is identical, the images are displayed correctly.

I’m baffled as to why the images fail to display when directly outputted by my GPT in the first instance, a phenomenon I’ve been able to replicate consistently. I’m hoping OpenAI could assist in resolving this issue to ensure that images in Markdown format are reliably and consistently displayed across all custom GPTs.

I’ve made a video documenting this issue ( It would be great if you could take a look and see if you’re able to replicate the problem I’ve described.


Saw similar issues in other posts Image Renderings in custom GPT's

1 Like

I’ve managed to resolve my issue, see How I Solved the GPT Image Display Glitch and Should We Keep the Security Mechanism? . I’m advocating for OpenAI to lift their restrictions, making similar issues more straightforward to address.

Hey folks, thanks for flagging this. We have resolved a few issues around generation URL’s. There are a few known edge cases that don’t work well:

  1. Having the model construct full URL’s on the fly (you should instead pass the full URL back in an Action response)
  2. Having the model fill in URL details (same as above, you should pass in the full URL and all UTM parameters)