Discover the Best AI Prompts for ChatGPT, MidJourney & More

By Perry Stevens, Blend Local Search Marketing | May 2026

The difference between mediocre AI output and expert-level results is almost never the tool — it is the prompt. The same ChatGPT, MidJourney, or Stable Diffusion model can produce generic fluff or brilliant, usable content depending entirely on what you feed it.

This guide gives you proven prompt frameworks and ready-to-use templates for the three most popular generative AI tools — so you get great results on the first try, every time.

Key Stat: Users who apply structured prompt frameworks (role + context + task + format) get 40-60% better quality output from ChatGPT compared to simple one-sentence prompts. For image generation, detailed prompts with style, lighting, and composition specifications produce 3x more usable images than vague descriptions. (Source: OpenAI Prompt Engineering Guide, 2025; MidJourney Community Research, 2024)

TL;DR

• ChatGPT: Use the Role-Context-Task-Format framework. Define who the AI is, what it knows, what you need, and how you want it delivered.
• MidJourney: Structure prompts as Subject + Style + Lighting + Composition + Parameters. Specificity beats length.
• Stable Diffusion: Use negative prompts extensively, add artist references for style control, and adjust CFG scale for prompt adherence.
• Bad prompts waste tokens and time. Good prompts compound — they produce output you can actually use without heavy editing.

Why Prompt Engineering Matters

AI models are probabilistic — they predict the next token (word, pixel, or concept) based on patterns in their training data. A vague prompt gives the model too much freedom, which usually results in generic, middle-of-the-road output. A structured, specific prompt constrains the probability space, guiding the model toward the exact result you want.

Think of it this way: asking ChatGPT "write a blog post" is like telling a chef "make food." Asking it to act as a seasoned local SEO consultant with 15 years of experience, writing for small business owners who are new to AI, in a conversational tone with bullet points and a TL;DR — that is a recipe the chef can actually execute.

ChatGPT Prompt Framework: The RCT Method

The most reliable framework for text-based AI is Role-Context-Task-Format (RCT). Every strong prompt contains these four elements:

1. Role

Define who the AI should act as. The more specific, the better.

• Weak: "You are a marketer."
• Strong: "You are a local SEO consultant with 15 years of experience helping small service businesses in competitive markets. You specialise in Google Business Profile optimisation and review generation strategies."

2. Context

Give the AI background information it needs to produce relevant output.

• Weak: "Write about SEO."
• Strong: "The reader is a plumber in Manchester who has a basic website but no Google Business Profile. He gets most customers through word-of-mouth and wants to start getting leads from Google Search. He has 2 hours per week to work on marketing."

3. Task

Be explicit about what you want the AI to do.

• Weak: "Help me with SEO."
• Strong: "Create a 4-week action plan for setting up and optimising his Google Business Profile, with specific tasks for each week, estimated time per task, and expected outcomes."

4. Format

Specify how the output should be structured.

• Weak: "Make it good."
• Strong: "Present the plan as a table with columns for Week, Task, Time Required, and Expected Outcome. Add a TL;DR summary at the top. Use UK English and a direct, practical tone."

Complete ChatGPT Prompt Template

Role: You are a [specific expert role] with [X years] of experience in [specific field].

Context: I am a [business type/audience description] in [location/market]. My current situation is [describe current state, constraints, goals].

Task: [Specific, actionable request — what you want produced]

Format: Deliver the output as [format: table / bullet list / email / blog post / script]. Include [specific elements: TL;DR, examples, step-by-step instructions]. Use [tone: professional / casual / persuasive] in [language variant: UK English / US English].

Constraints: [Optional: max length, things to avoid, must-include elements]

MidJourney Prompt Framework

MidJourney interprets prompts visually. The key is stacking descriptive modifiers that guide the aesthetic, not just the subject. Use this structure:

[Subject] + [Style/Medium] + [Lighting] + [Composition/Camera] + [Mood/Atmosphere] + [Parameters]

Example: Strong vs. Weak MidJourney Prompt

• Weak: "a dog"
• Strong: "a golden retriever puppy playing in autumn leaves, golden hour lighting, shallow depth of field, shot on Canon EOS R5, photorealistic, warm tones, cinematic composition --ar 3:2 --v 6"

MidJourney Modifier Categories

Category	Examples
Style/Medium	oil painting, digital art, photorealistic, anime, 3D render, watercolour, pixel art
Lighting	golden hour, softbox lighting, dramatic rim light, overcast, neon backlighting
Camera/Composition	wide-angle lens, macro shot, aerial view, Dutch angle, rule of thirds, shallow depth of field
Mood/Atmosphere	melancholic, energetic, mysterious, serene, cyberpunk, dystopian, whimsical
Artist Reference	in the style of Studio Ghibli, inspired by James Jean, reminiscent of Ansel Adams

MidJourney Parameters Quick Reference

• --ar 16:9 — Aspect ratio (landscape)
• --ar 9:16 — Aspect ratio (portrait, mobile)
• --v 6 — Model version (6 is latest)
• --s 250 — Stylisation (0-1000, higher = more artistic)
• --c 50 — Chaos (0-100, higher = more variation)
• --no text, blur, watermark — Negative prompt (exclude elements)

Stable Diffusion Prompt Framework

Stable Diffusion gives you more technical control but requires more precision. The most important technique is the negative prompt — telling the model what NOT to generate.

Stable Diffusion Prompt Structure

Positive prompt: [quality tags] + [subject] + [details] + [style] + [lighting] + [composition]
Negative prompt: [things to avoid] + [quality flaws] + [unwanted elements]

Example: Stable Diffusion Prompt

• Positive: "masterpiece, best quality, professional headshot, confident businesswoman, navy blazer, soft studio lighting, clean grey background, sharp focus, corporate portrait, 8k uhd"
• Negative: "blurry, low quality, deformed hands, extra fingers, cropped, out of frame, watermark, signature, text, bad anatomy, amateur"

Key Stable Diffusion Settings

Setting	What It Does	Recommended Range
CFG Scale	How strictly the model follows your prompt	7-12 (higher = more literal)
Steps	Iterations spent refining the image	20-50 (30 is usually enough)
Sampler	The algorithm used to generate	DPM++ 2M Karras or Euler a
Seed	Starting random number — same seed = reproducible results	-1 for random, or lock for consistency

Ready-to-Use Prompt Templates for Small Business

ChatGPT: Write a Service Page

Role: You are an experienced copywriter who specialises in local service business websites. You write clear, persuasive content that ranks well in search and converts visitors into leads.

Context: I run a [service business type] in [location]. We serve [target audience] and our main services are [list 3-5 services]. Our unique selling point is [what makes you different]. Most of our competitors have generic, boring websites.

Task: Write a 600-word service page for our main service: [service name]. The page needs to rank for "[primary keyword]" and related terms.

Format: Structure with: H1 title (include keyword naturally), introduction paragraph, 3 H2 sections covering benefits, process, and why choose us, a short FAQ section with 3 questions, and a call-to-action paragraph. Use UK English. Include the keyword in the first 100 words.

MidJourney: Product Mockup

professional product photography, [your product] on a clean marble surface, soft natural window light from the left, shallow depth of field, minimal shadows, premium brand aesthetic, neutral warm tones, shot on Sony A7IV 85mm f/1.8, photorealistic, 8k uhd --ar 4:5 --v 6

MidJourney: Social Media Hero Image

modern flat illustration, diverse team collaborating around a laptop, bright and optimistic colour palette, clean vector style, professional but friendly, SaaS marketing visual, ample negative space for text overlay, trending on Behance --ar 16:9 --v 6 --s 500

Stable Diffusion: Consistent Brand Character

Positive: "professional illustration, friendly mascot character, [describe appearance], consistent design, clean lines, flat colours, brand guidelines style, vector art aesthetic, white background"
Negative: "photorealistic, 3D render, messy, inconsistent style, blurry, low quality, watermark, text"
Settings: CFG 9, Steps 30, Seed locked for batch consistency

Common Prompt Mistakes (And How to Fix Them)

1. Being Too Vague

Fix: Add at least three specificity layers: subject, style, and context. "A logo" becomes "a minimalist line-art logo for an organic coffee brand, earthy green and brown palette, scalable vector style, no text."

2. Overloading the Prompt

Fix: MidJourney and Stable Diffusion have token limits. Prioritise the 3-5 most important modifiers. Test, then iterate — do not try to specify everything in one go.

3. Ignoring Negative Prompts (Stable Diffusion)

Fix: Always include a negative prompt. Start with the standard set: "blurry, low quality, deformed, extra limbs, watermark, text, signature, cropped" and add domain-specific exclusions.

4. Not Iterating

Fix: Treat your first result as a draft. Note what worked and what did not, then adjust 1-2 variables at a time. The best AI users are rapid iteration machines — not genius prompt writers.

About the Author

Perry Stevens is the founder of Blend Local Search Marketing, a Singapore-based agency that helps small businesses use AI tools for content, design, and automation. With over 15 years in digital marketing, he has tested hundreds of prompt frameworks across ChatGPT, MidJourney, and Stable Diffusion to find what actually produces usable business output — not just impressive demos.

FAQ

Do I need to be technical to write good AI prompts?

No. Good prompting is about clarity and structure, not coding. The RCT framework (Role-Context-Task-Format) works for anyone. Start with the templates above and adjust for your specific needs.

Which AI tool is best for my business?

It depends on what you need: ChatGPT for text (emails, blogs, code, analysis), MidJourney for high-quality marketing images and illustrations, Stable Diffusion for maximum control and privacy (runs locally, no subscription). Most businesses use a combination.

How do I get consistent results across multiple images?

In MidJourney, use --seed with the same number and similar prompts. In Stable Diffusion, lock the seed and use the same model/checkpoint. For brand consistency, create a style reference (a set of 4-6 images) and reference them in each prompt.

Are AI-generated images safe to use commercially?

Generally yes for most tools, but read the terms. MidJourney allows commercial use for paid subscribers. Stable Diffusion (open source) has no restrictions. ChatGPT-generated text is yours to use. Always check the current terms of service as they change.

What is the best free way to start with AI image generation?

Start with Bing Image Creator (powered by DALL-E 3, free with Microsoft account) or Leonardo.ai (free tier available). Both give you good results without subscription costs. When you outgrow them, upgrade to MidJourney or run Stable Diffusion locally.