New GPT-5.4 clobbers humans on pro-level work in OpenAI's tests - by 83%

OpenAI's release of GPT-5.4, which outperforms humans on pro-level work by 83% in the company's tests, marks a significant leap in AI capabilities. This new model is also more reliable, producing 18% fewer errors and 33% fewer false claims than its predecessor, GPT-5.2. For website owners, this means that AI-generated content may become increasingly prevalent and sophisticated, potentially blurring the lines between human-created and AI-generated material. As GPT-5.4 is capable of producing high-quality content with fewer errors, website owners may need to reassess their content strategies and consider the potential impact of AI-generated content on their sites. This could include evaluating the authenticity of user-generated content and developing strategies to detect and manage AI-generated material. To prepare for this shift, website owners can take several steps: first, review their content moderation policies to account for the increasing sophistication of AI-generated content; second, consider implementing AI-detection tools to identify and flag potentially AI-generated material; and third, update their llms.txt files to reflect the changing AI landscape and ensure that their sites are properly configured to handle AI-generated traffic.