A good prompt is less about secret keywords and more about giving an AI system the right job, relevant material and a clear definition of success.

Start with four building blocks

1 · TASK2 · CONTEXT3 · CONSTRAINTS4 · OUTPUT FORMAT

Specify what you need, who it is for, what to avoid, and the shape of the answer. If factual accuracy matters, tell the model what sources to use and where uncertainty should be disclosed.

Example: article research

EXAMPLE PROMPT · RESEARCH

Summarize the official release notes for this AI tool for a nontechnical reader. Explain what changed, include 3 concrete use cases, distinguish confirmed facts from interpretation, and link the original documentation. Avoid unverified claims.

Compare this with “Write about AI.” The detailed request gives the model a useful structure and quality criteria.

Example: photo editing

EXAMPLE PROMPT · IMAGE EDIT

Keep the same product, camera angle, shadows, and materials. Replace only the plain background with a warm ivory studio wall. Preserve label text exactly. Natural soft daylight, realistic reflections.

For image edits, describe what must stay the same as carefully as what should change.

Example: AI video

EXAMPLE PROMPT · IMAGE TO VIDEO

A slow cinematic push-in toward the ceramic cup. Morning sunlight moves gently across the table. Steam rises subtly; the cup and logo remain unchanged. Single continuous shot, realistic motion.

Video prompts benefit from describing motion, camera movement and continuity, rather than repeatedly describing an image already uploaded.

Iterate deliberately

Change one variable at a time: length, angle, wording, camera movement, or source image. Keep the successful pieces. The official OpenAI prompting best practices recommends clear instructions and iterative refinement.

Primary sources

Official vendor documentation was used to verify the workflow. Some options depend on plan, region, product version or account access.