OpenAI chief Sam Altman announced a significant update to ChatGPT on Friday, revealing that the AI chatbot should now follow user instructions to avoid using em dashes. Altman described this development as a small but happy win, sparking immediate discussion across social media platforms.
Mixed Reactions from ChatGPT Users
The announcement generated considerable interest among users who have long sought more control over the chatbot's writing style. While some celebrated the update as a step toward more customizable AI interactions, others expressed skepticism about whether the fix was functioning properly.
Simon Hedlin, an X user, highlighted how this change could benefit users who appreciate em dashes but want to avoid being accused of using unedited AI-generated content. His perspective reflected the cautious optimism among those who view punctuation preferences as subtle indicators of human authorship in the age of AI-generated text.
Users Report Continued Glitches
Despite OpenAI's official claim that the em dash issue had been resolved, several users shared evidence suggesting otherwise. Multiple X users posted screenshots demonstrating ChatGPT continuing to use the punctuation mark even after receiving explicit instructions to avoid it.
Ethan Dong described his experience as a big but disappointing loss, sharing visual proof of ChatGPT ignoring his directive. Another user, Dr Rishabh Jain, reported similar frustration, noting that despite asking the model to stop using em dashes, he received a response containing one. Jain called this peak betrayal, emphasizing the ongoing disappointment among users expecting more precise compliance from the AI.
Confusion Over Feature Activation
The conversation revealed that not all users understood how to activate the new setting. User @sandy_carter openly questioned how to enable the instruction feature, mentioning unsuccessful attempts to make it work.
Adding to the confusion were instances where ChatGPT itself used em dashes while responding to requests specifically asking it to avoid them. These occurrences highlighted the inconsistency that some users continue to experience with the supposedly updated feature.
When tested independently, the results mirrored user reports. After receiving the instruction not to use em dashes, ChatGPT responded with Got it — I'll avoid using em dashes from now on, ironically including the very punctuation mark it was being asked to avoid.
The mixed responses demonstrate the challenges in implementing consistent behavioral changes in AI systems, even for seemingly straightforward adjustments like punctuation preferences.
