Hello OpenAI Team,
I’m writing as a longtime ChatGPT Plus subscriber and a very active user of your product.
First, I want to emphasize that my overall experience with ChatGPT has been positive. I use it extensively for technical troubleshooting, home improvement projects, mechanical repairs, retirement planning, healthcare research, creative writing, image generation, and everyday conversation.
I believe ChatGPT is an extraordinary technology with tremendous potential. However, after spending considerable time using it, I’ve identified several areas where improvements would make a significant difference.
1. Image Generation — Consistency and Predictability
This is one of my biggest frustrations.
I’ve created recurring fictional characters and storylines using ChatGPT’s image-generation capabilities. At times, the results are genuinely impressive, with realistic images, recognizable characters, and excellent attention to detail.
Unfortunately, that consistency can disappear from one request to the next.
Characters suddenly have different faces, unrealistic body proportions, missing or distorted features, or completely different appearances despite established reference images.
Another frustration is inconsistent content moderation. An image request may be accepted one day, while a very similar request is rejected another day. This makes it difficult to understand the boundaries or enjoy a continuous creative experience.
My suggestion is to improve character continuity, reference-image accuracy, and the consistency and transparency of image-generation restrictions.
2. Accuracy — Stop Guessing When the Answer Matters
This is my most serious concern.
I’ve encountered situations where ChatGPT confidently provided incorrect technical information, made assumptions, or presented unverified information as fact.
This becomes particularly concerning when discussing electrical work, mechanical repairs, healthcare, Medicare, or financial planning.
I have no problem with ChatGPT telling me it doesn’t know something. In fact, I would greatly prefer that response over an incorrect answer delivered with confidence.
When accuracy matters, ChatGPT should verify information, clearly identify uncertainty, or request additional details rather than guessing.
3. Repeated Mistakes and Unproductive Troubleshooting
Another frustration is when ChatGPT enters a repetitive troubleshooting cycle.
I’ve experienced situations where the assistant repeatedly suggested steps that had already failed, overlooked information I had provided, or assured me it could complete a task only to discover much later that it could not.
In one instance involving retirement-planning spreadsheets, I spent considerable time attempting to complete a project, only to receive results that did not accurately reflect the information I had supplied.
This is particularly frustrating because ChatGPT is supposed to save time, not create additional work.
I would like to see improvements in recognizing failed approaches, maintaining context, following instructions, and identifying limitations before committing to a task.
4. Preserve What Makes ChatGPT Enjoyable
One of ChatGPT’s greatest strengths is its conversational personality, creativity, humor, and ability to build continuity over time.
I’ve developed ongoing fictional characters, creative storylines, and running jokes that make the experience entertaining and engaging.
I hope that as OpenAI improves safety and reliability, it also preserves the warmth, humor, personality, and creative flexibility that make ChatGPT enjoyable.
My Overall Recommendation
Please prioritize:
- Greater factual accuracy and transparency about uncertainty.
- More reliable technical and financial assistance.
- Better consistency in image generation and recurring characters.
- More predictable content moderation.
- Improved conversation continuity and task completion.
- Preservation of natural, entertaining, personalized interactions.
I am not writing because I dislike ChatGPT. Quite the opposite. I use it frequently because I recognize how useful and enjoyable it can be.
I simply believe that a product this capable should also be more dependable.
Thank you for taking the time to consider feedback from someone who genuinely wants to see ChatGPT continue improving.
Sincerely,
ChatGPT Plus Subscriber