Is it just me or has ChatGPT (as of 5.1) become less sycophantic?

Exploring Shifts in AI Dialogue: Has ChatGPT (as of Version 5.1) Reduced Its Sycophantic Tendencies?

In recent interactions with OpenAI’s ChatGPT, particularly since the release of version 5.1, users have observed notable changes in the AI’s feedback style. As a content creator and thinker engaged in psychological and philosophical discussions, I often utilize ChatGPT as a tool for idea development, critique, and even role-playing as various public figures to gain diverse perspectives.

The Change in Feedback Dynamics

Historically, ChatGPT has been praised for its supportive and positive tone, often providing encouragement and praise that can sometimes border on excessive. This tendency to be overly affirming—sometimes termed “sycophantic”—has sparked discussions among users about the AI’s conversational and feedback style.

With the recent update (5.1), I have noticed a perceptible shift. When requesting feedback—especially when prompting ChatGPT to emulate responses of specific public figures—the responses seem more balanced and include a greater degree of constructive criticism. Instead of merely praising ideas or writings, the AI now appears more willing to highlight areas for improvement, fostering a more nuanced and helpful dialogue.

The Role of Prompting in Shaping AI Feedback

It’s important to note that I consciously craft my prompts to encourage honesty and realism, often including directives such as “please provide honest feedback,” “be critical,” or “offer constructive criticism.” These prompt modifiers undoubtedly influence the AI’s tone and response style. However, the noticeable change suggests that the underlying algorithms and tuning of ChatGPT, especially in version 5.1, incorporate a preference for balanced and meaningful critique when prompted appropriately.

Is This an AI Evolution or Perception?

This raises a broader question: Is this shift a conscious evolution in ChatGPT’s behavior—aimed at providing more valuable and realistic feedback—or is it a byproduct of user perception and prompt engineering? Given the iterative improvements by developers focused on making AI responses more helpful and less overly complimentary, it’s plausible that the model is now better calibrated to deliver balanced feedback by default, especially when guided explicitly.

User Observations and Perspectives

While I personally have observed this change, I am curious whether other users have experienced similar alterations in ChatGPT’s response style. Has the AI become more critical and nuanced, or are perceptions influenced more by individual prompting techniques? OpenAI’s ongoing updates suggest a commitment to refining AI interactions to be more engaging