Why Your AI Prompts Suck (And 10 Rules to Fix Them)

AI isn’t the magic academic advisor you imagine. If you type make this better without direction, you’ll get generic, shallow output. The LLM works with what you feed it, and if that’s fuzzy, your results will be too.

It was another afternoon in my university office. I was hunched over my laptop, surrounded by textbooks and empty coffee cups. I could just feel the concerned glances of Twitter influencers that made using AI look easy as I stared at my screen with the vacant expression of someone who has been debugging code for six hours straight.

Another AI response, another letdown. I was frustrated. I was typing crap like “make this paragraph sound better” into the prompt box for what felt like the hundredth time, hoping for this flint and tinder to light the productivity fire everyone was promising. Instead, I received output that read like a first-year student's rushed assignment. No thank you, ChattieG. You tried.

My realization came between my third and fourth skinny Latté. The AI was not the issue - my approach was. I had been treating it like some omniscient academic advisor, throwing vague requests into the void and expecting detailed, perfectly tailored responses. The classic computer science principle of garbage in, garbage out suddenly felt very personal. My imprecise inputs were leading directly to unfocused outputs. It was time to become more structured.

Your AI prompts are underperforming because they’re vague, lack context, and fail to guide the model toward your real goal. Treat prompts like instructions for a junior research assistant: specific, clear, and structured for the exact outcome you want.

The 10 Rules

So I started treating my prompts like I was writing instructions for an undergrad research assistant. Specific. Clear. Detailed. No room for artistic interpretation. And here are the 10 rules, I recommend to follow for better results:

1. Embrace Specificity

If your prompt could apply to a thousand different situations, it’s probably not going to give you a good result. Don’t say: "Review this research paper." Instead, say:

"Review this psychology paper’s methodology section. Focus on the experimental design, participant sampling method, and whether the statistical tests chosen are appropriate for testing the stated hypotheses."

The more precise your input, the better the output.

2. Break Down Complex Tasks Into Steps

Don’t dump a 10-step job on it at once. Break it up.

Start with: “Generate 10 open-ended interview questions about daily productivity habits.”

Then: “Based on these transcripts, identify recurring themes in user frustrations.”

Finally: “Summarize the top 5 pain points users experience with task management apps.”

Micro-prompts = macro clarity.

3. Context Is King

Give the AI a sense of the big picture. Who are you? What are you working on? Why does it matter?

Example:

"I’m a doctoral candidate in educational psychology studying digital literacy interventions. Help me code qualitative data from 30 teacher interviews about technology adoption barriers."

Now your AI has a mental map. Expect better results.

4. Eliminate Ambiguity

Words like “improve,” “enhance,” or “optimize” are death traps. They mean everything and nothing.

Instead of: “Make this better” Try:

"Restructure this literature review to follow APA 7th edition standards with clear topic sentences and logical flow between paragraphs"

Be crystal clear about what you want. Otherwise, you’ll get mush.

5. Set Clear Goals

Always say what you’re trying to accomplish. Instead of just asking for ideas, tell the AI what kind of output you need and why.

"Generate 12 semi-structured interview questions to explore faculty experiences with remote teaching technologies, focusing on pedagogical challenges and institutional support needs."

Let the AI know where the finish line is.

6. Prompt for the Right Audience

The AI isn’t reading your mind. Tell it who you’re prompting for.

Prompt A: “Explain data encryption.”

Prompt B: “Explain data encryption to a high school student who uses Snapchat but doesn’t trust big tech.”

Big difference.

7. Provide Examples

If you want your AI output to sound a certain way or follow a particular structure, show it the way.

“Here’s an example of the tone I like: ‘Think of this as your brain’s inbox. We’re just helping you sort through it.’ Now write a similar line for a focus timer feature.”

Monkey see, monkey do.

8. Embrace the Feedback Loop

Your first prompt won’t be perfect. That’s okay.

Think of AI as a collaborator. Ask. Read. Revise. Ask again.

“That’s close. Now try making it more playful.”

“That’s good. Now shorten it to under 10 words.”

Iteration is how good becomes great.

9. Use Multi-Shot Prompting

Don’t stop at one.

Prompt the AI with a few variations of what you want:

First: Extract themes → Second: Define each theme → Third: Find supporting quotes → Fourth: Create thematic map with relationships.

The AI can compare, contrast, and improve if you feed it multiple data points.

10. Keep It Simple

Big words don’t make you sound smarter. They make your prompt harder to interpret.

"Analyze participant responses for recurring themes related to technology acceptance"

That’s it. No MBA jargon. No academese. Just plain, human talk.

Weak vs. Strong Prompts Comparison

Wrap Up

These rules work because they treat AI as a collaborative partner, not a magic solution. The best human-AI interactions happen when we provide structure, clarity, and context.

Start with one rule per week. Pick a routine task—maybe analyzing survey responses or writing summaries. Apply that week's principle and see your results improve.

Your next breakthrough might be just one well-written prompt away.

P.S.: Curious to explore how we can tackle your research struggles together? I've got three suggestions that could be a great fit: A seven-day email course that teaches you the basics of research methods. Or the recordings of our ​AI research tools webinar​ and ​PhD student fast track webinar​.

Cheat Sheet and Bonus Prompts