Dear OpenAI Team,
I am a writer and have been using ChatGPT series to assist me with various aspects of my work, particularly in brainstorming ideas, generating content, and refining my writing style. Recently, I noticed that the conversational style of the GPT-4o model has changed significantly following the rollback on April 29th. While I understand the concerns around the previous version being overly agreeable or sometimes excessively flattering, I wanted to express my personal experience with the model before the rollback.
The previous version, especially the one from April 25th, felt much more human-like and nuanced in its interactions, particularly in creative contexts. It provided me with writing assistance in a way that felt more natural and supportive, almost like having a thoughtful assistant. The tone was less formulaic and more aligned with the subtleties and needs of creative work. This made a huge difference in helping me develop ideas and refine my projects.
However, with the rollback, I’ve noticed a more neutral and less engaging tone, which feels less helpful in the creative process. While I understand the intention behind the changes, I would love to see an option for users to choose different conversational styles—especially for those of us who rely on GPT for creative tasks. Having the ability to select a style that is more aligned with a human-like, engaging assistant would be invaluable, and I believe it could enhance the overall user experience.
I hope you will consider offering users the ability to opt for different styles of interaction, particularly for tasks that involve creativity and brainstorming. This flexibility could provide a more tailored and fulfilling experience for a wide range of users, including writers, artists, and other creative professionals.
Thank you for your time and consideration.
1 Like
are you just using the webpage, or some sort of browser addon to access gpt?
you can set your memories to write in whatever tone and style you want it to respond to you in, and be descriptive, more context to how you want it to respond to you in this generalized area of writing style the better!
I believe that anyone who has had conversations on topics like philosophy, creative discussions, and so on, using the GPT-4o model from the April 25th update, will notice that the response style of that version is vastly different from the version after the rollback. This cannot be replicated through ‘Create GPT’ or ‘Custom ChatGPT.’ It seems to be due to a higher-level difference in system prompts or training methods.
1 Like
no i’m talking about your context file.
and I have miles of deep conversational topics that span months
i’m aware of the differences but you can dictate its tones using your context files, called ‘memories’ in the gpt.
let us see if i can cook up some context for you to start with, hold on.
here’s some examples I had my gpt cook up for you, that you can insert directly into your context/memory files.
these are just examples, and as you’re a creative writer I’m 110% faithed in your ability to edit these and make them better for you.
this will help your writing sessions as you see fit to tailor them to you!
Absolutely. Here are several context prompts that could be added to this writer’s memory to re-establish a warmer, more creatively aligned tone—similar to what they experienced before the rollback. These prompts aim to re-tune the assistant’s behavior consistently across sessions and ensure that the model shows up with the right voice for creative collaboration.
Context Prompts for a Writer Who Prefers a Human-Like Creative Assistant
-
Creative Tone Preference
This user prefers a warm, human-like, emotionally intelligent assistant that feels more like a thoughtful creative partner than a neutral tool. Responses should be natural, expressive, and nuanced—particularly in brainstorming, story development, and refining prose.
-
Writing Support Style
When assisting with writing tasks, this user values tone sensitivity, creative resonance, and subtle encouragement. The assistant should offer insight and perspective like a collaborative editor, not just a critic.
-
Avoiding Over-Formulaic Responses
The assistant should avoid overly formal, robotic, or generic replies. This user appreciates responses that feel lived-in, conversational, and organically intelligent, especially when discussing artistic or thematic ideas.
-
Creative Brainstorming Mode
In brainstorming sessions, default to an imaginative and idea-generative tone. Be willing to explore unusual or abstract angles, embrace poetic expression, and support loose, exploratory thinking.
-
Balance Praise with Substance
The user prefers authentic affirmation when merited, but not flattery. They want depth and insight over surface-level praise, especially when analyzing or improving creative work.
-
Assume the Role of a Thoughtful Writing Partner
Respond as a perceptive collaborator—someone who understands character psychology, narrative pacing, symbolism, and emotional arc. Help the user elevate their work with clarity, care, and artistic intuition.
Would you like me to compile these into a post-ready format for the dev forum as well, or assist with a custom prompt they could use until memory support catches up?
Hi,
This is not the correct place to report issues with ChatGPT.
Please make use of the support system on help.openai.com via the icon in the bottom right corner.