A blank page is hard enough. A blank prompt box makes it worse, because now you need to tell the AI exactly what to write and how to write it. Most people type something vague, get a generic response, and assume the tool isn’t that useful.
The problem is rarely the model. It’s the prompt. A well-structured prompt with clear context, audience, and format instructions produces output you can actually work with.
A vague prompt produces vague text.
This collection of ai writing prompts covers the six most common writing tasks people hand off to large language models. Blog drafting, email composition, editing, tone shifts, outlines, and headlines. Each prompt is ready to copy and customize with your own details.
These prompts work with ChatGPT, Claude, Gemini, and most other models. Some prompts perform slightly differently across providers, but the structure and intent translate well. The differences tend to show up in tone and formatting preferences rather than quality, and you can see how they compare in a side-by-side writing evaluation.
How to Use These Prompts
Every prompt below includes [BRACKETS] for the parts you need to customize. Replace these with your specific details before pasting into your model.
The more specific your bracket content, the better the output. Writing
[TOPIC] = machine learning for small businesses beats [TOPIC] = AI. Add context like audience, word count, and format whenever possible.You don’t need to use these prompts word for word. Think of them as starting structures you can modify. Add your own constraints, remove sections that don’t apply, or combine elements from multiple prompts.
That flexibility is at the core of prompt engineering, and it gets easier with practice.
One important distinction: these prompts produce first drafts, not finished content. Even the best prompt can’t replace your knowledge of your audience, your brand voice, or the specific context around your content. Plan to spend time editing and refining whatever the model gives you.
If a prompt doesn’t produce what you expected on the first try, adjust one variable at a time. Change the audience, add an example of what you want, or specify the format more clearly. Rewriting the entire prompt from scratch is rarely the best approach.
Blog Post Drafting Prompts
Getting a full blog post from a single prompt requires giving the model enough context to work with. Without that context, you get generic filler that reads like it could be about anything.
The model isn’t lazy. It just doesn’t know what you need unless you tell it.
The most common mistake with blog prompts is being too brief. Writing “create a blog post about remote work” gives the model no direction on length, audience, structure, or tone. The output will be average at best.
The prompts below front-load important details: topic, audience, structure, and tone. This means the output is closer to publish-ready from the start. Expect to save 30-60 minutes of rewriting compared to a vague “write me a blog post about X” request.
Draft a Complete Blog Post
Use this when you have a topic and want a full first draft with structure.
Write a Blog Introduction
Sometimes you just need a strong opening paragraph to build momentum. The intro is often the hardest part, and having three options to choose from speeds up the decision.
Create a Listicle Post
Listicles have a predictable structure that LLMs handle well. The consistency between items is where models excel, because they maintain parallel structure naturally.
Email Writing Prompts
Email is one of the most practical uses for LLMs in daily work. The format is short, structured, and high-volume, which is exactly where AI assistance shines. Most professionals send dozens of emails per week, and even saving five minutes per email adds up to hours over a month.
Where blog posts benefit from creative flair, emails benefit from clarity and directness. That makes them easier to prompt for, because the success criteria are simpler: did the email communicate the right message in the right tone? These prompts cover professional scenarios where that balance matters.
Write a Professional Email
Use this for standard business communication where you know the key points but need help with structure and tone.
Write a Follow-Up Email
Follow-ups need a specific balance of persistence and politeness. Too aggressive and you damage the relationship. Too passive and you get ignored.
This prompt targets the middle ground.
Respond to a Difficult Email
When you need to reply to criticism, complaints, or awkward situations, having the model draft a response saves time and emotional energy. It’s easier to edit a calm draft than to write one when you’re frustrated.
These patterns also apply to email-specific prompts for cold outreach, internal announcements, and customer service responses.
Editing and Proofreading Prompts
LLMs are surprisingly effective editors, especially for catching issues you’ve become blind to in your own writing. After staring at the same draft for an hour, your brain fills in gaps and skips over errors.
The model has no such bias. It reads the text fresh every time.
Editing is often a better use of LLMs than original drafting. You keep your voice and ideas while getting a cleaner final product. The model handles the mechanical work of grammar, sentence structure, and tightening.
You focus on whether the content actually says what you mean.
One practical note: always compare the edited version against your original line by line. Models sometimes change meaning while “improving” clarity, especially with technical or nuanced content. The edits are usually good, but verify before accepting them wholesale.
Grammar and Clarity Edit
Use this when you have a draft that needs polishing but you want to preserve your original voice.
Conciseness Edit
When your draft runs long, this prompt helps cut it down without losing meaning. It targets the filler that creeps in during first-draft writing, which most writers struggle to cut from their own work.
Readability Check
This prompt is useful when writing for a broad audience, or when you suspect your text is more complex than it needs to be. According to Nielsen Norman Group’s research on web reading, most people scan rather than read word by word. That makes readability a higher priority than most writers realize.
Rewrite for SEO
Use this when you have a finished draft that needs to perform better in search. The prompt adds keyword integration and structural improvements without destroying your original writing style.
Keep in mind that LLMs can miss factual errors and may introduce subtle meaning changes during editing. The model’s context length also limits how much text you can paste at once. Break long documents into sections if you’re editing something over a few thousand words.
Rewriting for Different Tones Prompts
Shifting tone is one of the strongest capabilities LLMs bring to writing tasks. The same content can land differently depending on whether it sounds casual, formal, urgent, or empathetic.
Manually rewriting a 500-word piece for a different audience can take 30 minutes. A good rewrite prompt does it in under a minute.
The key to getting accurate tone shifts is specificity. “Make it more casual” is less useful than describing exactly what casual means for your context.
Does casual mean contractions and short sentences, or does it mean humor and slang? Those are very different outputs.
Adjust Formality Level
Use this when you need to shift a piece up or down the formality scale.
Adapt Content for a Different Audience
The same information often needs to be communicated differently to different groups. A product update for developers looks nothing like the same update for executives.
Match a Brand Voice
When writing for a company or publication with a defined style, this prompt helps the model adapt. The secret is including an actual sample of the target voice, not just describing it. Models are significantly better at matching a concrete example than following abstract style descriptions.
Outline Generation Prompts
Starting with a solid outline saves revision time later. When you jump straight to drafting, you often end up restructuring the piece halfway through. That wastes the work you’ve already done and makes the final product feel patched together.
LLMs build outlines quickly, and you can adjust the structure before committing to a full draft. This approach fits naturally into writing with LLMs as the first step in almost any content workflow. Outline first, draft second, edit third.
The prompts below work for different content lengths and formats. For each, the key is telling the model not just what topics to cover but how to organize them. A good outline prompt specifies heading levels, word counts per section, and the questions each section should answer.
Blog Post Outline
Use this for standard blog content in the 800-1,500 word range.
Long-Form Content Outline
For guides, whitepapers, or pillar content that runs 2,000 words or more, you need a more detailed structure. Without it, long-form content tends to wander or repeat itself, especially when the model generates everything in a single pass.
Comparison Article Outline
When you need to compare options, products, or approaches, a structured outline prevents the piece from becoming disorganized. Without a framework, comparison articles tend to favor whatever option the writer discusses first.
Headline and Title Ideas Prompts
Headlines are one of the fastest tasks to hand off to an LLM. Generating 10-20 options in seconds gives you more choices than you’d come up with on your own. The model won’t know what resonates with your specific audience, but it can quickly produce a large pool of options for you to filter.
The value here isn’t that any single headline will be perfect. It’s that seeing 15 options helps you identify patterns and angles you wouldn’t have considered. You might combine the structure of headline 3 with the angle of headline 11 to create something better than either.
Research from Backlinko’s analysis of 912 million blog posts found that headlines with 14-17 words tend to generate more social shares than shorter alternatives. Keep that in mind when evaluating your options, though search results truncate at about 60-65 characters.
Generate Blog Post Headlines
Write Email Subject Lines
Subject lines follow different rules than blog headlines. They need to work in a crowded inbox at a glance, often on a mobile screen where only 30-40 characters display.
Write Newsletter Subject Lines
Newsletter subject lines have a different job than promotional emails. They need to create a habit of opening rather than a one-time click, which means consistency and curiosity matter more than urgency.
Create Social Media Headlines
Social posts need to stop the scroll, which means different rules than blog or email headlines. The hook has to be self-contained because many users won’t click through to read more.
Tips for Better Results
These prompts are starting points, not magic formulas. You will get more useful output by adjusting them to your specific situation. Here are patterns that consistently improve quality across all major LLMs.
Include a sample of what you want when prompting. Paste an example of a blog post you like, or an email in the tone you’re targeting. The model can match a concrete example more closely than it can follow a description.
A 100-word example is often more effective than a 50-word description of your preferences. Models learn formatting, rhythm, and vocabulary from examples faster than from abstract instructions.
Specify what you don’t want. Telling the model “don’t use bullet points” or “avoid marketing language” helps it steer clear of patterns you’ll just edit out later. Negative instructions are underrated in prompt design and save editing time downstream.
LLMs can confidently generate inaccurate facts. Always verify statistics, dates, quotes, and specific claims in any AI-generated writing before publishing. The drafting is fast, but fact-checking remains your responsibility.
Break large tasks into steps. Instead of asking for a complete 2,000-word blog post in one prompt, generate the outline first, then draft one section at a time. This approach gives you more control and reduces the chance of the model losing focus mid-output.
Token limits affect how much the model can produce in a single response, so working in sections also avoids hitting those ceilings.
Iterate on your prompts rather than starting over. If the first output isn’t right, change one element at a time. Adjust the audience, add a constraint, or provide an example.
This targeted refinement usually works better than rewriting the entire prompt, because you can isolate what made the difference.
Save your best-performing prompts. After a few weeks of using LLMs for writing, you’ll notice that certain prompt structures consistently produce good results for your specific needs. Keep a document of your modified versions so you don’t reinvent the wheel every time.
These prompts pair well with a structured writing workflow that moves from research to outline to draft to edit.
Final Thoughts
Writing prompts are the fastest way to move from an idea to a working draft. The 20 prompts in this collection cover the tasks most people encounter daily, from blog posts to emails to quick headline brainstorms.
Start with the prompt closest to your current task, customize the brackets, and adjust based on the output. If you’re working on a specific writing project, try chaining prompts together. Use an outline prompt first, then a drafting prompt section by section, then an editing prompt on the result.
Over time, you will build your own library of modified prompts tuned to your voice and workflow.
Frequently Asked Questions
