2. Go to ChatGPT. com and add "/" click on "Create image" After that, upload your product image + add this prompt: "Create a high-resolution fashion photoshoot of a female model wearing the exact olive green ribbed bodycon dress shown in the reference image. The model has a
PROMPT ENGINEERING
-

Reasoning models achieve top performance without extended thinking
By
–
So apparently… Reasoning models can be effective without thinking! This flips the table on assumptions that long, detailed chains of thought are always necessary. Turns out, with the right training methods, even concise responses can deliver top-tier performance.
-
Using delimiters to separate instructions from content
By
–
11/ Use delimiters to separate instructions from content. When pasting text into Windsurf or Cursor, wrap it like: Your text goes here… Then add: "Summarize the content above in 3 paragraphs." This prevents confusion about what's what. BONUS: When using
-
Provide examples for complex outputs across apps
By
–
10/ For complex outputs across apps, provide examples. GPT-4.1 follows examples precisely. Show exactly what you want:
"Answer in this format:
SUMMARY: One sentence
ANALYSIS: Three paragraphs
RECOMMENDATION: Bullet points" -
Provide clear format for creative tasks with GPT-4.1
By
–
8/ For creative tasks, give it a clear FORMAT. GPT-4.1 excels at following structural guides: "Write a story with:
Setting: [x]
Characters: [x]
Plot points: [x]
Theme: [x]
Length: [x]" Using this approach in tools like Windsurf / Cursor gives better creative output. -
Precision over volume: clear instructions for GPT-4.1
By
–
6/ When it's not doing what you want, don't repeat yourself louder – be more precise. GPT-4.1 doesn't need ALL-CAPS or "please please please." It needs clear, unambiguous instructions.
"Write in short paragraphs of 2-3 sentences each" works better than "MAKE IT SIMPLE!!" -
Use chain of thought for research and analysis tasks
By
–
7/ Use "chain of thought" for research or analysis tasks. Ask GPT-4.1 to:
– Break down what you're asking
– Consider key factors
– Analyze options
– Reach a conclusion This structure produces dramatically better results than just asking for an answer. -
Prompt “think step by step” improves GPT-4.1 accuracy
By
–
4/ Ask it to "think step by step" before answering complex questions. GPT-4.1 needs explicit permission to show its reasoning. Adding this phrase improves accuracy by ~4%. Example prompt:
"First, think carefully step by step about what documents are needed to answer the query. -
Place important instructions at the beginning and end of the prompt
By
–
3/ Put important instructions TWICE – at the beginning AND end of your prompt. GPT-4.1 has a 1M token context (4x larger than before), but this means your key instructions can get lost. Bookend them! "I need a formal tone… [rest of request]… Remember, keep it formal."
-

Structure of system prompts: persistence, tools, planning
By
–

2/ Structure system prompts with these 3 main elements – Persistence: Tell the model it's in a multi-turn conversation. It shouldn’t stop until the user's question is fully answered. – Tool Use: Encourage the model to use tools instead of guessing. – Planning: Let the model
