Five prompt habits that produce bad output

Even with the best frameworks and techniques, certain habits consistently produce poor results. Recognising these anti-patterns in your own behaviour is half the battle. Once you spot them, the fixes are straightforward.

Anti-Pattern 1: The Vague Dump

What it looks like:

Write me something about marketing.

Why it fails: This prompt gives the AI almost nothing to work with. "Something" could be a blog post, a strategy document, a social media calendar, or a haiku. "Marketing" could refer to any of hundreds of sub-disciplines. The AI will produce a response, but it will be a generic guess — and you will waste time either rewriting it from scratch or trying to steer it towards what you actually needed.

The fix: Add specifics. Every time you write a prompt, ask yourself: Could ten different people interpret this in ten different ways? If yes, you need to narrow it down.

Fixed version:

Write a 300-word blog post introduction explaining why small e-commerce businesses should prioritise email marketing over social media advertising, aimed at solo founders with limited marketing budgets. Use a conversational, encouraging tone.

Anti-Pattern 2: The Kitchen Sink

What it looks like:

Write me a business plan, also create a marketing strategy, and come up with a name for the company, and design the customer persona, and suggest pricing.

Why it fails: When you cram multiple unrelated tasks into a single prompt, the AI gives shallow treatment to each one. It is like asking someone to simultaneously cook dinner, do the laundry, and help your child with homework — everything gets half-done.

The fix: One task at a time. If you have five things to accomplish, have five focused interactions (or five clearly separated steps within one conversation).

Fixed version:

Let us work through my business plan step by step. We will tackle each section in a separate message. First, let us start with the customer persona. My product is a project management app designed for freelance graphic designers. Help me build a detailed persona for my primary customer segment, including demographics, pain points, goals, and typical workflow.

Anti-Pattern 3: The Mind Reader

What it looks like:

Make this better.

(Attached: a document with no context about what "better" means to the person.)

Why it fails: "Better" is entirely subjective. More concise? More persuasive? More formal? Better structured? More evidence-based? The AI cannot read your mind, and it will default to generic "improvements" that may not align with what you actually wanted.

The fix: Be explicit about your criteria. Tell the AI what dimension of "better" you care about.

Fixed version:

Review this project proposal and improve it in three specific ways: (1) make the executive summary more compelling by leading with the business impact, (2) tighten the language throughout — cut any sentences that do not add new information, and (3) add a risk mitigation section after the timeline. Keep the overall length under 1,500 words.

Anti-Pattern 4: The One-Shot Wonder

What it looks like: You send a prompt, receive a response, and immediately copy-paste it into your final deliverable without any iteration.

Why it fails: AI's first response is a first draft. Sometimes first drafts are excellent. More often, they are 70–80% of the way there and need refinement. By accepting the first response every time, you are leaving significant quality on the table.

The fix: Treat every response as a starting point. Read it critically. Ask yourself: What is missing? What is weak? What would I change if I were editing this? Then tell the AI.

Fixed version approach:

(After receiving the first response)

This is a solid start. Three changes: (1) the opening paragraph is too generic — start with the client's specific challenge instead, (2) the recommendations section needs concrete numbers, not just qualitative statements, and (3) the tone in the final paragraph shifts to being too casual — keep it consistent with the rest.

Anti-Pattern 5: The Over-Engineer

What it looks like:

You are an expert marketing strategist with 20 years of experience in B2B SaaS technology companies serving the enterprise segment, particularly those in the financial services vertical. Your communication style should be incisive yet approachable, data-driven but not dry, and you should balance strategic vision with tactical pragmatism. When responding, first consider the macro-economic landscape, then narrow to industry-specific trends, then focus on the specific tactical recommendation. Format your response using H2 headers for each section, bullet points for key takeaways, and bold text for critical terms. Ensure your response accounts for both short-term wins and long-term brand building... (continues for another 300 words)

...What subject line should I use for this email?

Why it fails: There is a mismatch between the complexity of the prompt and the simplicity of the task. Over-engineering wastes your time, can confuse the AI by burying the actual request in excessive context, and creates diminishing returns. A 500-word prompt for a task that needs a 50-word prompt is not thoroughness — it is inefficiency.

The fix: Match the effort in your prompt to the complexity of the task. A quick question deserves a quick prompt. A complex, high-stakes deliverable deserves a detailed prompt. Use your judgement.

Fixed version:

Suggest five email subject lines for a cold outreach email to CFOs at mid-size UK fintechs. We are selling an expense management platform. Make the subject lines concise, curiosity-driven, and professional.

The Common Thread

Notice the pattern across all five anti-patterns: they are all failures of calibration. The Vague Dump gives too little information. The Kitchen Sink asks for too much at once. The Mind Reader assumes too much. The One-Shot Wonder stops too early. The Over-Engineer tries too hard.

Effective AI communication is about finding the right level of specificity, scope, explicitness, iteration, and effort for each individual task. That calibration instinct develops with practice — and the exercises that follow will give you exactly that.