AI From Zero · Prompt Engineering

Common Prompting Mistakes

Learn to identify and avoid common prompting mistakes that lead to unclear, incomplete, inconsistent, or unreliable AI responses.

Estimated learning time: 20 minutes

What You'll Learn

By the end of this lesson, you will be able to:

- Identify common mistakes people make when writing AI prompts.
- Recognize vague and ambiguous instructions.
- Explain why missing context can reduce response quality.
- Identify conflicting instructions and unnecessary complexity.
- Understand the problems caused by missing output requirements.
- Recognize unrealistic or contradictory constraints.
- Improve prompts by making requirements specific and consistent.
- Identify when examples or additional context are needed.
- Avoid relying on AI assumptions when important information is missing.
- Apply a systematic approach to improving weak prompts.

Why Prompting Mistakes Matter

A prompt can fail even when the AI system is capable of performing the requested task.

Often, the problem is not the AI itself but the way the task has been communicated.

A prompt may be too vague, lack important context, contain conflicting requirements, or fail to explain what the desired result should look like.

Mistake 1: Being Too Vague

One of the most common mistakes is giving an instruction that is too broad.

For example:

Write something about AI.

The AI has to make many decisions about the subject, audience, purpose, length, and format.

A clearer prompt would define what aspect of AI is required and who the content is for.

Mistake 2: Not Defining the Goal

A prompt should make the intended outcome clear.

For example:

Tell me about customer service.

could mean an overview, training material, improvement recommendations, or a research summary.

Instead, specify the goal:

Identify five practical ways a small business can improve customer service response times.

Mistake 3: Missing Context

AI may produce a technically reasonable response that is unsuitable because important context was not provided.

For example, a recommendation for a large company may not work for a small business.

Provide the context that materially affects the task.

Mistake 4: Not Defining the Audience

The same information may need to be explained differently depending on who will use it.

For example:

Explain machine learning to a complete beginner.

is more specific than simply asking:

Explain machine learning.

Mistake 5: Not Specifying the Output Format

If the format matters, state it.

For example:

Compare these three options in a table with columns for cost, benefits, risks, and recommendation.

Without the format requirement, the AI may return a long paragraph instead.

Mistake 6: Using Ambiguous Language

Words such as quickly, professional, simple, or detailed can have different meanings.

When precision matters, explain what you mean.

For example:

Keep the response under 300 words and use short paragraphs.

This is clearer than:

Keep it short.

Mistake 7: Giving Conflicting Instructions

Two requirements can sometimes contradict one another.

For example:

Provide a comprehensive explanation but use no more than 50 words.

The AI has to balance incompatible requirements.

Review the prompt to ensure that its major instructions can reasonably be satisfied together.

Mistake 8: Too Many Unnecessary Instructions

A prompt can become unnecessarily complicated by including instructions that do not affect the result.

More instructions do not automatically mean better prompting.

Include requirements that are relevant to the actual task.

Mistake 9: Hiding the Important Requirement

If one requirement is particularly important, make it explicit.

For example:

Most importantly, preserve all financial figures exactly as provided.

This makes the priority clear.

Mistake 10: Not Specifying What to Preserve

When rewriting or transforming content, important information can accidentally be changed or removed.

Specify what must remain unchanged.

For example:

Rewrite the email in a friendlier tone but preserve all dates, amounts, and commitments.

Mistake 11: Not Specifying What to Exclude

Sometimes the AI needs to know what should not appear in the response.

For example:

Focus only on the supplied report and do not introduce external assumptions.

Mistake 12: Asking for Too Much in One Unstructured Prompt

A complex task can become difficult to follow when many unrelated requests are combined without structure.

Instead, organize the task into clear sections or stages.

Mistake 13: No Clear Sequence for Multi-Step Tasks

If a task contains several stages, specify the order.

For example:

  1. Summarize the report.
  2. Identify the three main risks.
  3. Compare the risks by potential impact.
  4. Provide recommended actions.

This is clearer than combining all four requests into one vague instruction.

Mistake 14: Asking for a Result Without Providing Necessary Information

AI cannot reliably use information that has not been provided or made available to it.

If the task depends on specific data, documents, or background information, provide the necessary material.

Mistake 15: Expecting AI to Guess Missing Information

When important information is missing, guessing can produce an unsuitable answer.

Instead, tell the AI what to do when information is unavailable.

For example:

If the owner or deadline is not provided, write Not specified rather than guessing.

Mistake 16: Not Defining the Desired Level of Detail

A response can be technically correct but still unsuitable because it is too detailed or too brief.

Specify the expected level.

For example:

Explain the concept in approximately 300 words using two practical examples.

Mistake 17: Using Vague Requests for Improvement

Requests such as make it better do not identify what should improve.

A stronger instruction identifies the problem.

For example:

The explanation is accurate but too technical. Rewrite it for beginners using plain language and one everyday example.

Mistake 18: Changing Too Many Things During Iteration

When refining a prompt, changing many unrelated elements at once makes it difficult to understand what caused an improvement or deterioration.

When practical, change one important dimension at a time.

Mistake 19: Failing to Review the Response

A good prompt does not eliminate the need to review the result.

Check whether the response actually satisfies the requirements.

Mistake 20: Assuming AI Output Is Automatically Correct

An AI response can sound confident while still containing errors.

Important facts, figures, dates, and conclusions should be verified when accuracy matters.

Mistake 21: Not Asking for Uncertainty

Some questions do not have a simple or certain answer.

You can ask AI to identify uncertainty.

For example:

Identify which conclusions are strongly supported and which remain uncertain.

Mistake 22: Asking for Unsupported Conclusions

A conclusion should be connected to the available evidence.

Instead of asking the AI to force a particular conclusion, ask it to evaluate the evidence.

Mistake 23: Not Providing Evaluation Criteria

If you ask AI to select the best option, it needs to know what best means.

For example:

Rank these options based on cost, ease of implementation, expected benefit, and risk.

Mistake 24: Using Examples Without Explaining Their Purpose

An example can help guide AI, but it should be clear what the example demonstrates.

For example:

Use this example to understand the desired structure and tone, then create a new response for the information below.

Mistake 25: Giving an Example That Conflicts With the Instructions

If the example suggests one format while the written instructions request another, the AI may have difficulty determining what you want.

Ensure examples and instructions are consistent.

Mistake 26: Over-Constraining Creative Tasks

Creative tasks need direction, but excessive restrictions can leave little room for useful variation.

Specify the important boundaries while allowing appropriate creative freedom.

Mistake 27: Under-Constraining Structured Tasks

The opposite problem can occur when a task needs a particular format but no structure is provided.

If you need a table, checklist, comparison, or sequence, state it clearly.

Mistake 28: Mixing Different Audiences

A prompt can become unclear if it simultaneously asks for content for very different audiences.

When necessary, create separate versions for each audience.

Mistake 29: Ignoring the Intended Medium

Content for an email, presentation, social media post, report, and video script may require different structures.

Specify where the content will be used.

Mistake 30: Not Defining Length Constraints When They Matter

If the content must fit within a particular space or time, specify the limit.

For example:

Write a 30-second video script.

or:

Keep the final message under 150 words.

Mistake 31: Forgetting Important Constraints

Before submitting a prompt, check whether there are requirements related to length, format, audience, tone, source material, or exclusions.

Mistake 32: Asking for Precision Without Defining the Criteria

If you want a precise comparison or evaluation, explain how precision should be judged.

For example:

Compare the options using the same five criteria and explain each score.

Mistake 33: Failing to Separate Facts From Opinions

When analyzing a subject, make it clear whether the response should distinguish factual information from interpretation or opinion.

Mistake 34: Asking for a Recommendation Too Early

For complex decisions, it can be useful to analyze the evidence before requesting a final recommendation.

For example:

  1. Identify the relevant factors.
  2. Compare the options.
  3. Explain the trade-offs.
  4. Then provide a recommendation.

Mistake 35: Not Challenging the Initial Answer

For important analytical tasks, consider asking the AI to examine its conclusion from another perspective.

For example:

Identify the strongest argument against your recommendation.

Mistake 36: Treating the First Prompt as Final

A useful prompt can be improved after seeing the first response.

Review what worked and what did not, then refine the prompt.

Mistake 37: Repeating the Same Prompt Without Learning From the Result

If the response is not useful, simply repeating the same prompt may not solve the underlying problem.

Identify what is missing and modify the instruction accordingly.

Mistake 38: Using Random Prompt Changes

Prompt improvement should be deliberate.

Change the instruction because you have identified a specific problem, not simply because the previous response was unsatisfactory.

Mistake 39: Not Preserving What Already Works

When refining a prompt, do not unnecessarily remove useful requirements.

Tell AI which parts of the existing result should remain unchanged.

Mistake 40: Failing to Define Success

A prompt becomes easier to evaluate when you know what a successful result looks like.

For example:

A successful answer should identify three major risks, provide evidence for each, and end with practical mitigation steps.

A Simple Prompt Review Checklist

Before submitting an important prompt, ask:

  • Is the goal clear?
  • Is enough context provided?
  • Is the audience clear?
  • Is the desired output format specified?
  • Is the required level of detail clear?
  • Are important constraints included?
  • Are the instructions consistent?
  • Have unnecessary instructions been removed?
  • Have important facts or source materials been provided?
  • Have I explained what should be preserved or excluded?
  • Do I know what a successful result should look like?

How to Fix a Weak Prompt

A practical improvement process is:

  1. Identify the problem with the current prompt.
  2. Determine what information or instruction is missing.
  3. Add the missing requirement.
  4. Remove conflicting or unnecessary instructions.
  5. Run the revised prompt.
  6. Evaluate the new response.

Example of a Weak Prompt

Write something about AI for my company.

This prompt does not define the audience, purpose, format, topic, or length.

Improved Prompt

Write a 500-word introductory article about practical uses of AI for small businesses. The audience is business owners with little technical knowledge. Use plain language, include five practical examples, and finish with three simple steps for getting started.

The improved prompt provides a goal, audience, length, topic, style, structure, and desired ending.

Another Example

Weak:

Analyze these numbers and tell me what is happening.

Improved:

Analyze the supplied sales data. Identify the three largest changes compared with the previous period, explain the patterns visible in the data, identify any unusual values, and present the findings in a table followed by a short summary.

Prompting Is a Communication Skill

Many prompting mistakes are really communication problems.

The better you communicate the goal, context, constraints, and expected result, the easier it becomes for AI to produce a useful response.

Do Not Make Prompts Complicated Just to Make Them Detailed

A strong prompt is not necessarily a long prompt.

The objective is to provide the information and instructions that matter.

Use Structure When Structure Helps

For complex tasks, headings, numbered steps, bullet points, and explicit output requirements can make the prompt easier to follow.

Review Before Sending

A few seconds spent checking the prompt can prevent multiple rounds of correction later.

Look especially for missing context, ambiguous language, conflicting requirements, and undefined output expectations.

What You Should Remember

  • Vague prompts often produce vague or unfocused responses.
  • Clear goals and relevant context improve task alignment.
  • Define the audience and desired output when they matter.
  • Avoid ambiguous or conflicting instructions.
  • Do not include unnecessary complexity.
  • Specify what information should be preserved or excluded.
  • Do not expect AI to reliably guess important missing information.
  • Use evaluation criteria when asking AI to compare or rank options.
  • Review and refine prompts based on observed problems.
  • Always review important AI output rather than assuming it is automatically correct.

What Comes Next?

The next lesson will build on these principles by focusing on how to improve prompt quality systematically and create more reliable prompting workflows.

Key Takeaways

  • Most prompting mistakes are communication problems that can be corrected with clearer instructions.
  • Important context, audience, goals, constraints, and output requirements should be explicit.
  • Conflicting or unnecessary instructions should be removed.
  • AI should not be expected to guess critical missing information.
  • Prompt refinement should be deliberate and based on observed problems.
  • Important AI output should always be reviewed.

Try It Yourself

Diagnose and improve a weak prompt.

Write a deliberately weak prompt for a task such as writing, research, analysis, or creative work.

Then identify at least five problems with it.

Rewrite the prompt by adding the missing goal, context, audience, constraints, evaluation criteria, and output format where appropriate.

Run both prompts and compare the results. Identify which changes produced the greatest improvement.

Test Your Knowledge

You've reached the end of this lesson.

Test what you've learned with the Common Prompting Mistakes - Quiz.

Take the Quiz
← Prompting for Research and Analysis
Advanced Prompting Strategies →
Back to Course