What is a prompt in AI?
The basic definition
In AI, a prompt is the instruction or query you provide to a generative model, such as a large language model (like ChatGPT) or an image generator (like DALL·E). It's the starting point that the model uses to predict what comes next—whether that's a sentence, a paragraph, a translation, or a picture.
Prompts can be as simple as a single word ("Hello") or as complex as a multi-paragraph brief with examples, constraints, and formatting rules. The model doesn't "understand" in a human sense; it generates output by statistically matching patterns it learned during training.
What a prompt can include
A prompt can contain several elements: a task description ("Summarize this article"), context ("for a busy executive"), examples ("here's a sample summary"), and output specifications ("use bullet points, max 100 words"). For image models, prompts often include subject, style, lighting, and composition details.
The model treats everything in the prompt as part of its input, so the order and wording matter. Clear, specific prompts tend to produce more useful results, but even vague prompts can work if the model has enough context from training.
- Task: what you want done (e.g., "Write a poem")
- Context: background or audience (e.g., "for a 5-year-old")
- Examples: samples of desired output
- Format: structure or length (e.g., "in a table")
- Constraints: rules or limits (e.g., "avoid jargon")
Common mistakes
- Thinking a prompt is only a question—it can also be a command, example, or partial text to complete.
- Assuming the AI understands your intent perfectly; it only sees the words you provide, not your thoughts.
- Believing a longer prompt is always better; irrelevant details can confuse the model.
