I’m using the API via XMLHTTP in VBA. The response is saved as raw data in Microsoft Access. I then display it in an Edge Web Control by converting everything to HTML in a stylish presentation, thanks to the Markdown text. This worked perfectly until last week’s ChatGPT update. That update, which strips all the Markdown from the response, broke my fancy display unless I chose not to use 5th generation models. It took a lot of work to achieve this level of aesthetics. Imagine you’ve done everything right on your side, and it looks like OpenAI quietly changed the rules and broke your entire pipeline without warning. Where’s the decency in warning the developers beforehand? They could add a parameter to specify whether the response is expected with or without Markdown. Now I’m furious and refuse to be held accountable for all these display issues because of OpenAI. Ironically, ChatGPT is helping me a lot in finding workarounds to fix its creator’s mistakes.
ChatGPT is the consumer chatbot platform. You are using the API.
Go back to the model you were using before and see. You are likely the one that made the model ID switch to something new.
You will find that o3 and gpt-5 are purposely post-trained on not producing markdown, and you have to be even more deliberate in your description with gpt-5 than simply “Response: formatted as markdown”
Yes indeed. It takes a lot more than “formatted as markdown”. Actually, it takes a permanent system message of 10 lines to force gpt-5 forgetting its “mission” to send a pure RTF text which is absolutely useless in a JSON message as it becomes plain text.