Prompt: meaning and practical example
A prompt is input given to an AI system, potentially containing a question, task, background, example or output requirements. It need not be one sentence or purely text, depending on the system. A precise prompt clarifies the task without guaranteeing a correct answer.
Also known as: AI prompt · instruction
At a glance
- Make the task, relevant context and desired outcome clear.
- Information supplied in a prompt does not automatically become a verified fact.
- More words are not always better; relevant detail and review matter.
On this page

What counts as a prompt?
A question to an AI is a prompt, as is a request to compare text or create a table. It can contain both instructions and material to process. Distinguish what the system should do from the document it should work on.
Google Cloud's overview describes prompts as input directing model behavior. Detailed instructions do not make a large language model infallible. Task clarity and factual verification remain separate.
What can a useful task description contain?
State the action, necessary background, important constraints and output form. Simple questions may need only one clear sentence rather than a long template.
| Part | Editing example |
|---|---|
| Task | Summarize the supplied service description. |
| Audience | Small-business owners making an initial inquiry. |
| Constraint | Do not add unverified services. |
| Format | A short introduction and three clear points. |
Do not demand certainty or detail unsupported by available data. Asking the system to flag missing information can be better than requiring every blank to be filled.
Example: clarifying a vague request
Imagine a bicycle repairer asking, “Write good copy.” The destination, audience and valid service claims are unspecified, encouraging generic output unrelated to the workshop.
A clearer request is: “Use the attached service list to introduce the start-of-season inspection. Rely on the listed work and do not invent prices or turnaround times.” Review is still needed, but the task and permitted sources are clearer. This is an illustration, not a performance test.
How does it differ from a search query?
A search query generally seeks information or results. Prompts can request rewriting, classification, comparison and explanation. The concepts overlap in AI search, but not every prompt starts web retrieval.
In conversational search, later prompts can refine earlier answers. Identify which detail should change. If goals or conditions changed, do not assume an ambiguous reference communicates that reliably.
What can detailed prompting not solve?
Better instructions do not make missing or incorrect sources true. AI hallucination remains possible even when the response confidently follows the requested format. Check claims, names, numbers and references.
Long contradictory instructions can also cause confusion. The context window is finite, and copied details are not equally relevant. Supply decision-critical information and separate distinct tasks.
How should the next attempt improve?
Identify the actual mismatch: wrong audience, missing criterion or unsupported claim. Request that specific correction while naming what to retain, instead of merely asking for something “better.”
A reusable short brief can help recurring work, but current information still needs checking. Prompts support rather than replace knowledgeable editing.
For visibility research, prompt tracking repeats defined observations, while a prompt set identifies which customer questions the results represent.