ChatGPT attempts to predict the best pattern of words to respond to any prompt. Sometimes it will believe that the best pattern is shorter than you asked for, either because it didn’t fully understand the prompt, or because the prompt wasn’t something it could fulfil.
It isn’t particularly strong at worrying about an exact length, as that wasn’t something it was trained for. It was trained to recognize and generate the best patterns, not specific length copy. You may need to break down longer generative tasks into smaller sub-sections or sub-processes.
It’s still all about an AI reacting to your prompts and input. You can’t change the AI, so you have to focus on the thing you can change - your prompts. Learning how to prompt the AI is what everyone is doing, and why AIPRM, giving access to thousands of pre-made, pre-tested prompts, is so useful to hundreds of thousands of users right now.
If you want to specify exactly what you are trying to achieve, what outcome you are looking for, in as much precise detail as possible, and want to share the prompts you have been using, then others may be able to suggest things to try, or what works for them. That is exactly why having a community forum is so valuable, after all.