Common prompting mistakes

Learn from everyone else's mistakes

Almost everyone makes the same handful of mistakes when they start using AI assistants. The good news: each one has a simple fix. Recognize these patterns, and you'll skip months of frustration.

Mistake 1: Being too vague

The most common mistake by far. Vague prompts get generic answers because the AI fills every gap with the most average option.

Fix: Ask yourself, "What would a smart stranger need to know to do this well?" Then include it.

Mistake 2: Cramming too much into one prompt

"Research my competitors, write a business plan, design a logo concept, and create a 12-month marketing calendar" is four big projects. The AI will do all of them shallowly.

Fix: Break big jobs into steps. Do one part, review it, then use the result in the next prompt. You'll get better quality at every stage and catch problems early. (This approach has a name you'll meet later: prompt chaining.)

Mistake 3: Hiding the real goal

People often ask for something adjacent to what they actually need. "Make this paragraph sound smarter" when the real goal is "convince my manager to approve the budget." The AI can only optimize for what you tell it.

Fix: State the purpose. "Revise this paragraph so it persuades my manager to approve a $5,000 software budget. She cares most about time saved."

Mistake 4: Not specifying the format

You wanted a quick checklist; you got five paragraphs. You wanted a table to paste into a spreadsheet; you got prose.

Fix: Say exactly what shape you need: "a numbered list," "a two-column table," "three bullet points under 15 words each," "plain text with no headings."

Mistake 5: Only saying what NOT to do

"Don't be too formal. Don't make it long. Don't use clichés." The AI now knows three things to avoid and nothing about what you want.

Fix: Turn each "don't" into a "do." Keep a restriction when it's genuinely important, but pair it with a positive direction.

Mistake 6: Accepting the first answer

Many people treat the first response as final: either it's good enough or "AI doesn't work for this." But the first answer is a draft.

Fix: Give feedback like you would to a colleague. Be specific about what to change and what to keep:

This is close. Keep the structure and the opening line, but make the middle section half as long and replace the generic examples with ones about small coffee shops.

Mistake 7: Trusting without verifying

The answer looks polished, so it goes straight into the report, the email, or the code. Then someone spots the invented statistic or the bug.

Fix: Match your checking to the stakes. Verify names, numbers, dates, quotes, and sources. Test code before relying on it. Remember the limitations lesson: fluent isn't the same as accurate.

A real-world example

Priya needed to prepare for a job interview. Her first prompt was "Help me prepare for an interview." She got a generic list of tips she'd read a hundred times (Mistake 1). Her second attempt asked for a full prep guide, mock questions, salary negotiation scripts, and a thank-you email all at once, and everything came back thin (Mistake 2).

Her third attempt worked:

I have a second-round interview on Thursday for a customer success manager role at a mid-sized software company. The interviewer is the head of support. I have three years of experience in retail management but I'm new to software. Give me the 6 questions I'm most likely to be asked about switching industries, and for each one, a short note on what the interviewer is really trying to find out. Don't write my answers yet; I'll draft those next.

Then she drafted her answers and asked for feedback on each one: one step at a time, with clear context, a clear task, and a clear format.

A quick pre-send checklist

Before you hit enter on anything important, run through:

1. Goal: Did I say what I'm trying to achieve?
2. Context: Did I include what the AI can't know?
3. Task: Is the request specific and appropriately sized?
4. Format: Did I describe what the answer should look like?
5. Safety: Did I remove anything sensitive?

And after you get the answer: refine, then verify.

Quick recap

Most prompting problems come from a few predictable mistakes: vagueness, overloading, hidden goals, missing format, negative-only instructions, accepting the first draft, and skipping verification. Each has a simple fix, and together they form the habits of a skilled prompter.