ChatGPT-5 is OpenAI’s most steerable model to date. This means it’s easier to guide it to an ideal output, but only if you understand Context engineering.
Get the Most Out of GPT-5
Be Explicit and Demanding: Tell the AI to "think harder" or "think deeper." This can trigger higher-quality reasoning.
Enforce Planning: For complex requests, instruct the AI to create a plan before it starts generating a response. This ensures a more structured and thorough output.
Structure is Key: Use clear and structured prompts. While formats like JSON can be helpful, the main benefit comes from the clarity and detail you provide.
Demand Transparency: Ask the AI to share its thought process. This can lead to more intelligent and well-reasoned answers.
Avoid Contradictions: Make sure your instructions are not conflicting, as this can confuse the model and degrade the quality of the response.
Encourage Self-Correction: Ask the model to create its own "excellence rubric" and then iterate on its response until it meets that standard.
Use Metaprompting: Ask the AI to help you improve your prompts. It can often provide valuable suggestions from its own perspective.
