Better AI results come from better instructions, and most beginners can get there without any technical vocabulary. A beginner gets more useful answers from ChatGPT, Claude, or Gemini by stating the task, adding context and an audience, naming the output format, setting a few constraints, and then refining the first draft with a follow-up message. The 25 prompts below apply that pattern to five everyday jobs, with editable placeholders in square brackets.
Generative AI answers the request it receives, and a vague request gets a vague answer. The five sets of prompts here cover emails, summaries, brainstorming, image generation, and video scripts.
Each set comes with a short before-and-after example showing how the answer changes. Readers can copy any prompt, swap in their own details, and send it.
They can then use the refinement habits from the first section to turn a decent draft into one they would put their name on.
What Makes a Prompt Work Better?
A prompt works better when it removes guesswork. Every piece of missing information (who the reader is, how long the answer should be, what it should look like) is a gap the model fills with a generic default.
Clear task statements, a defined output format, and a willingness to revise cover most of what beginners need.
Start with a Clear Task, Context, and Audience
The first line of a prompt should name the job with a concrete verb: draft, summarize, list, compare, rewrite. “Help me with an email” leaves the model guessing.
“Draft a reply declining a meeting” tells it exactly what to produce. Context comes next.
This covers the situation, the relationship, and any facts the model cannot know, such as a deadline, a product name, or what happened in the last conversation. Audience shapes tone and vocabulary more than almost anything else.
A summary for a CEO and a summary for a new intern should read differently, and the model adjusts once it knows who will read the result. A widely shared illustration shows the difference.
The weak prompt “Write about marketing” becomes “Write a 300-word beginner-friendly explanation of digital marketing trends in 2026 using simple language,” and that better prompt example from KnowledgeHut adds a length, an audience, a topic, and a style in one sentence.
Specify the Output Format and Useful Constraints
Output format tells the model what shape the answer should take. Common options include:
- Bullet points or a numbered list
- A short email with a subject line
- Headings with one paragraph each
- A two-column comparison written as paired bullets
- Scene-by-scene script blocks
Asking for structured output saves editing time, because the answer arrives ready to paste. Constraints narrow the result further.
Word limits, reading level, and tone all count, and so do negative constraints such as “no jargon” or “avoid exclamation marks.” In image tools, the same idea is called negative prompting: listing what should stay out of the picture.
One caution about long conversations. Every model has a context window, a limit on how much text it can hold at once.
In very long chats, early instructions can lose their grip, so restating key constraints in a later message keeps the output on track.
Improve a Vague Prompt Through Iterative Refinement
Few first drafts are final, and treating the first answer as a starting point is one of the fastest ways to improve. Guidance from major AI developers points the same way: clear instructions, useful context, and iterative testing beat clever wording.
Here is a before-and-after sequence:
- Before: “Write a thank-you note to a client.”
- Result: A stiff, generic paragraph that could apply to anyone.
- After: “Write a 90-word thank-you note to [client name], who signed a six-month contract with our design studio yesterday. Keep the tone warm and professional, mention our kickoff call on [date], and end with one sentence about what we’re excited to build together.”
- Result: A short note with a specific date and a personal close, ready to send after a quick read.
When a result misses, a follow-up message fixes it faster than starting a new chat. Useful follow-ups include “Make it shorter,” “Use simpler words,” “Sound less formal,” or “Keep paragraph two, rewrite the rest.”
Several common prompt engineering mistakes show up again and again:
- Leaving out the audience
- Asking for several unrelated tasks in one prompt
- Accepting the first draft without review
- Copying a template without replacing the placeholders
Copy-Paste Prompts for Emails and Summaries
Email and summary prompts improve most when they include the reader, the goal, and a length limit. The ten prompt templates below cover drafting, rewriting, replies, and condensing long material into something scannable.
Email Prompts for Drafting, Rewriting, and Replies
A strong email prompt answers four questions: who is receiving it, what the sender wants them to do, how it should sound, and how long it should be. Subject lines deserve extra attention, since 47% of recipients open emails based on the subject line alone.
Before-and-after example
- Before: “Write a follow-up email.”
- After: “Write a friendly 80-word follow-up to [name], who hasn’t replied to my proposal from [date]. Remind them of the main benefit, [benefit], and ask if a 15-minute call next week works.”
- Why it helps: The model now knows the relationship, the history, the goal, and the length, so the draft reads like a real follow-up.
Prompt 1: Draft a new email
Write an email to [recipient and their role] about [topic]. The goal is to get them to [desired action]. Use a [friendly / professional / direct] tone. Keep it under [number] words and end with one clear next step. Include three subject line options under 50 characters.
Prompt 2: Rewrite a rough draft
Rewrite the email below so it sounds [more confident / warmer / more concise]. Keep all facts, names, and dates exactly as written. Cut filler phrases and keep it under [number] words. Email: [paste draft]
Prompt 3: Reply to a difficult message
Draft a calm, professional reply to the message below. Acknowledge the concern about [issue], explain [what I can offer], and avoid sounding defensive. Keep it under 150 words. Message: [paste message]
Prompt 4: Decline politely
Write a short email declining [request or invitation] from [name]. Thank them, give a brief reason ([reason]), and suggest [alternative, if any]. Tone: kind and firm. Maximum 100 words.
Prompt 5: Create a follow-up sequence
Write two follow-up emails for [name], who showed interest in [product/service] but stopped responding. Email one goes out after [number] days and is a gentle check-in. Email two goes out after [number] days and offers [helpful resource or incentive]. Keep each under 90 words and avoid pushy language.
Summary Prompts for Articles, Meetings, and Research
Summary prompts work best when they say who the summary is for and what the reader needs to do with it. A manager wants decisions and deadlines; a student wants main ideas and definitions.
Before-and-after example
- Before: “Summarize this meeting.”
- After: “Summarize these meeting notes for a teammate who missed the call. List decisions made, open questions, and action items with owners and due dates. Use bullet points and keep it under 200 words.”
- Why it helps: Naming the reader and the three categories turns a vague recap into a document someone can act on.
Prompt 6: Summarize an article
Summarize the article below in [number] bullet points for [audience]. Start with a one-sentence main takeaway. Flag any claims that include statistics so I can verify them. Article: [paste text]
Prompt 7: Turn meeting notes into action items
Turn these meeting notes into three sections: Decisions, Action Items (with owner and due date), and Open Questions. If an owner or date is missing, write “not assigned.” Notes: [paste notes]
Prompt 8: Condense research for a decision
I’m deciding whether to [decision]. Summarize the research below into: the three strongest points for, the three strongest points against, and what information is still missing. Use plain language for a non-expert. Research: [paste text]
Prompt 9: Explain a dense document simply
Explain the document below as if to [audience, e.g., a new employee]. Use short paragraphs, define any technical term the first time it appears, and keep the whole explanation under [number] words. Document: [paste text]
Prompt 10: Compare two sources
Compare the two texts below. List where they agree, where they disagree, and which points only one of them covers. Keep each list to five items or fewer. Text A: [paste] Text B: [paste]
A practical habit with every summary: spot-check at least one or two points against the original. Natural language processing tools condense text quickly, but they sometimes blur numbers or merge separate ideas.
Copy-Paste Prompts for Brainstorming
Brainstorming prompts produce sharper ideas when they name the audience, set a quantity, and push the model away from the obvious. Role prompting, sample ideas, and requests for reasoning each give the model a clearer target.
Generate Ideas for a Specific Audience
Asking for “ideas” gets the first ideas anyone would think of. Asking for ideas aimed at a named group, with a quantity and a ban on clichés, gets a list worth reading.
Before-and-after example
- Before: “Give me blog post ideas about fitness.”
- After: “Act as a content strategist. Give me 10 blog post ideas for busy parents who want to exercise at home in under 20 minutes. Skip generic topics like ‘benefits of exercise.’ For each idea, add one sentence on why this audience would click.”
- Why it helps: The role sets a perspective, the audience narrows the topic, and the ban on generic picks forces fresher angles.
Prompt 11: Audience-specific ideas
Act as a [role, e.g., marketing strategist]. Generate [number] ideas for [project or content type] aimed at [specific audience and their main problem]. Avoid the most obvious suggestions. For each idea, add one line explaining why it fits this audience.
Prompt 12: Unexpected angles
List [number] ideas for [topic]. Make the first three conventional, then make the rest progressively more unusual. Label each idea as “safe,” “fresh,” or “bold.”
Compare Options and Solve a Problem
Once ideas exist, the next job is picking one. Asking the model to show its reasoning makes the comparison easier to check and easier to challenge.
Prompt 13: Weigh options
I’m choosing between [option A], [option B], and [option C] for [goal]. My priorities are [priority 1] and [priority 2], and my budget is [amount]. Compare the options on each priority, explain your reasoning step by step, and then recommend one.
Prompt 14: Solve a stuck problem
Here’s a problem I’m stuck on: [describe problem]. I’ve already tried [what was tried]. Suggest five possible causes and, for each one, a small test I could run this week to check it.
The “already tried” detail matters in practice. Without it, the model spends half its answer on the fixes the user ruled out an hour ago.
Use Examples to Shape More Relevant Ideas
Few-shot prompting means showing the model a few samples of what good looks like before asking for more. This is one of the most dependable prompt engineering techniques for matching a voice or format, because examples communicate tone faster than descriptions do.
Prompt 15: Ideas modeled on examples
Here are three [headlines / product names / taglines] I like: [example 1], [example 2], [example 3]. Generate [number] more in the same style for [new topic]. Match the length and tone, and don’t reuse any words from my examples.
The final instruction keeps the new ideas from becoming near-copies of the samples, which is the most common problem with example-based prompts.
Copy-Paste Prompts for Image Generation
Image prompts succeed when they describe what the viewer should see: subject, setting, lighting, style, and framing. AI models that generate images, including tools in the Gemini and ChatGPT family, respond to concrete visual words far better than to adjectives like “beautiful” or “cool.”
Describe a Scene with a Clear Subject and Setting
The subject goes first, followed by what it is doing and where. One clear subject produces cleaner results than a crowded list of competing elements.
Before-and-after example
- Before: “A cozy coffee shop.”
- After: “A small corner coffee shop on a rainy evening, a barista pouring latte art at a wooden counter, warm yellow pendant lights, rain streaking the front window, shot from a customer’s seat.”
- Why it helps: The model gets a subject, an action, a setting, lighting, and a viewpoint, so it stops inventing them at random.
Prompt 16: Basic scene
Create an image of [subject] [doing what] in [setting] at [time of day]. The mood should feel [mood]. Show it from [viewpoint, e.g., eye level, above, close-up].
Prompt 17: Product shot
Create a clean product photo of [product] on [surface] with a [color] background. Use soft natural light from the [left/right], a slight shadow beneath the product, and leave empty space on the [side] for text.
Set the Style, Composition, and Visual Details
Style and composition details turn a generic image into a specific one. One useful checklist for visual prompts groups subject, action, background, lighting, camera angle, style, and mood into a single structured prompt.
Visual details worth specifying include:
- Art style (watercolor, flat illustration, realistic photo, 3D render)
- Lens or framing (wide shot, 85mm portrait, overhead)
- Color palette (muted earth tones, bright primary colors)
- Aspect ratio (square for feeds, vertical for stories, wide for banners)
Prompt 18: Styled illustration
Create a [art style] illustration of [subject] in [setting]. Use a [color palette] palette, [lighting description], and a [composition, e.g., centered subject with lots of negative space]. Aspect ratio: [ratio].
Prompt 19: Image with negative prompting
Create a [style] image of [subject and setting]. Do not include [unwanted elements, e.g., text, extra people, watermarks, cluttered background]. Keep the focus on [main subject].
Negative prompting works best with a short, specific list. Long exclusion lists sometimes pull the unwanted element into the image anyway, so three or four exclusions is a sensible limit.
Revise an Image Prompt Without Starting Over
Most chat-based image tools accept edits in plain language, so a near-miss image rarely needs a fresh prompt. Changing one thing at a time makes it clear which edit caused which result.
Prompt 20: Targeted revision
Keep this image the same, but change [one element, e.g., the background to a sunset beach]. Don’t change the subject’s pose, clothing, or the lighting direction.
Good revision requests name what to keep as well as what to change. “Make it better” gives the model nothing to work with, while “make the sky darker and keep everything else” produces a focused edit.
Copy-Paste Prompts for Video Scripts
Video script prompts work when they state the platform, the length in seconds, the viewer, and the single action the viewer should take at the end. ChatGPT, Claude, and other AI language models can draft a full script in one pass once those details are clear.
Turn an Idea into a Short Video Script
A short script needs a hook in the first few seconds, a small number of points, and a clear ending. Stating the length in seconds keeps LLMs from writing a three-minute script for a 30-second clip.
Before-and-after example
- Before: “Write a video script about saving money.”
- After: “Write a 45-second vertical video script for young adults who just got their first job. Open with a surprising hook in the first three seconds, give three simple saving tips, and end by asking viewers to comment with their own tip. Mark the hook, each tip, and the ending clearly.”
- Why it helps: Length, audience, structure, and ending are all defined, so the draft fits the platform instead of reading like a blog post read aloud.
Prompt 21: Idea to script
Write a [length in seconds]-second video script about [topic] for [audience]. Start with a hook in the first three seconds. Cover [number] key points in plain language. End with [call to action]. Label each part: Hook, Point 1, Point 2, and so on, and Close.
Adapt a Script for Different Audiences and Platforms
A script that works on one platform often needs trimming or reshaping for another. Asking the model to adapt an existing script saves time and keeps the core message consistent.
Prompt 22: Platform adaptation
Adapt the script below for [platform, e.g., YouTube Shorts / LinkedIn / TikTok]. Adjust the length to [seconds], match the typical tone of that platform, and keep the main message about [message]. Script: [paste script]
Prompt 23: Audience adaptation
Rewrite this script for [new audience, e.g., retirees / high school students / small business owners]. Change the examples and vocabulary to fit them, but keep the same structure and length. Script: [paste script]
For creators who make similar videos every week, saving standing instructions (brand voice, banned phrases, preferred length) as custom instructions or system prompts means they don’t have to be retyped each time.
Request Scenes, Voiceover, and a Call to Action
A production-ready script separates what viewers hear from what they see. Requesting both columns of information up front makes filming or editing much faster.
Prompt 24: Scenes plus voiceover
Write a [length]-second script for [topic] with each scene as its own block. For each scene, include: Visual (what’s on screen), Voiceover (exact words), On-screen text (short caption), and Duration in seconds. Total length must add up to [length] seconds.
Prompt 25: Stronger call to action
Here is the ending of my video script: [paste ending]. Write five alternative calls to action that ask viewers to [desired action]. Keep each under 12 words, make them specific, and avoid phrases like “smash that like button.”
Reading the voiceover aloud once before recording catches lines that look fine on screen but trip up the speaker.
Make Each Prompt Your Own
These 25 prompts become effective prompts only after the placeholders hold real details: a named reader, a real deadline, an actual product. The more specific the filled-in version, the less editing the answer needs.
A simple routine keeps results improving:
- Fill every bracket. Leftover placeholders produce generic output.
- Read the first draft as a draft. Mark what works and what misses.
- Send one targeted follow-up. Ask for a single change, such as shorter, warmer, or a different format.
- Save the winners. When a filled-in prompt produces a great result, store it in a notes file for next time.
- Review before using. Check facts, names, and numbers against the original material.
Iterative refinement gets faster with practice. Most beginners can learn foundational prompting skills within a few weeks through consistent use, and the fastest route is picking one prompt from each category above and running it on a real task this week.