Prompting basics
What are the most common beginner prompting mistakes?
Most beginner trouble comes from vague requests, missing context, no example format, and telling the model what to avoid instead of what to do. Each has a direct fix.
Most early frustration traces back to a few habits. The good news is each one has a direct fix that the earlier pages already cover.
Being vague. “Write something about our product” leaves every real decision to the model. Be specific about the output you want and its constraints, which is the first thing both Anthropic and OpenAI recommend.[1, 2] Name the length, the audience, and the goal.
Giving no context. The model does not know who the output is for or what it will be used for. Anthropic notes that supplying the motivation behind an instruction helps the model deliver a more targeted response.[1] A sentence of background usually beats another adjective.
Not showing a format. If the shape of the answer matters, show an example of it. Examples are among the most reliable ways to steer format and structure, and a handful of input/output examples goes a long way.[1, 2]
Saying what not to do. A list of bans leaves the model guessing what you do want. Anthropic’s advice is to tell the model what to do instead: rather than “Do not use markdown,” try “Your response should be composed of smoothly flowing prose paragraphs.”[1] A positive instruction points at a target; a prohibition only fences off one path.
Expecting one perfect try. Treat the first answer as a draft and refine from there. Prompting is iterative, and the second prompt is usually the better one.[2]
One more assumption is worth naming even though it does not fit neatly into the five above: assuming shared context that was never actually given. “Fix the bug we talked about” or “use the file I mentioned” means nothing to a model that only sees what is literally present in this conversation. There is no memory of a different chat, no visibility into a browser tab, no access to a file that was not pasted in or attached. If it is not in the prompt, the conversation history, or an attached file, treat it as invisible to the model, no matter how recently you looked at it yourself.
References
- Prompting best practices — Anthropic
- Prompt engineering — OpenAI