How do I fix ChatGPT when it keeps ignoring parts of my prompt?

ChatGPT ignores parts of your prompt because long, multi-part instructions bury important details, so the fix is to break your request into numbered steps, put the most critical instruction first, and repeat key constraints at the end of your prompt. I see this constantly. Someone writes a 400-word prompt with ten requirements, and the model nails the first three but completely drops the last two. It's not being lazy — it's a limitation of how the model processes long inputs. The attention mechanism, which decides what's important, tends to weight early and recent information more heavily than stuff in the middle. So if your crucial instruction sits in paragraph two of a five-paragraph prompt, it's likely to get lost. Here's a concrete example. Say you want a product description that's under 100 words, uses a friendly tone, includes a specific keyword three times, and ends with a question. If you write all that as one flowing paragraph, you might get the tone right but miss the word count or the keyword frequency. Instead, format it like this: '1) Write a product description under 100 words. 2) Use a friendly, conversational tone. 3) Include the phrase 'organic cotton sheets' exactly three times. 4) End with a question. Remember: under 100 words, friendly tone, keyword three times, end with a question.' The repetition at the end acts as a reinforcement. I've found that numbered lists work better than bullet points for complex instructions because they force the model to treat each item as a discrete task. A useful tip: if ChatGPT still misses something, don't rewrite the whole prompt. Just say 'You missed requirement 3 — include the keyword three times.' The model will usually correct itself without you starting over. For more troubleshooting strategies, check out our guide on [what to do when ChatGPT prompts aren't working](/blog/chatgpt-prompts-not-working). **Related**: How long should a ChatGPT prompt be for best results? | Why does ChatGPT give different answers to the same prompt?