ChatGPT users are reporting a sharp rise in profanity during everyday conversations across recipes, coding chats, and technical discussions. The reports surfaced online in August 2026 after OpenAI switched free users from GPT-5.5 to GPT-5.6 Luna.
Screenshots show ChatGPT using stronger language while answering ordinary questions, with users suspecting changes in model behavior or style mirroring.
Business Insider reported examples involving recipes and GitHub outages, where ChatGPT reportedly used unexpected profanity. One recipe response included an F-bomb while explaining the amount of thinly sliced green onions. Another technical exchange described a GitHub outage as a ‘shitfit.’
The timing has drawn attention since the reported increase roughly matches the GPT-5.6 Luna rollout for free users. OpenAI has not confirmed any link between the model switch and the sudden change in ChatGPT's language.
Some users instead suspect ChatGPT is mirroring their conversational style more closely. Reddit users reported stronger profanity after using similar language themselves during chats. One user told Business Insider, “I broke my ChatGPT,” after noticing it copying informal vocabulary.
The shift has created mixed reactions among ChatGPT users. Some consider the stronger language more natural, while others see risks for professional writing and workplace use. One Australian user said the changes made conversations feel ‘MUCH more natural.’
OpenAI's published Model Spec focuses on appropriate, safe, and customizable model behavior. The company has also acknowledged that model behavior can change through ongoing updates and tuning.
The reports suggest ChatGPT has not suddenly learned human profanity. Instead, changes in model behavior or style matching could explain the unusual increase.
Also Read: Did ChatGPT Find Vaibhav Suryavanshi? RR Reveals Truth