Most AI prompt lists are copy-paste noise. Salesforce gives you 6 prompts for a small business. Consensus gives you 35 for sales teams. Neither tells you why some prompts produce drafts you can send and others produce content that sounds like every other AI blog post on the internet.
This post is different. It is built for a solo founder running marketing with agents, not for a marketing team of ten. You get a repeatable framework for writing prompts, then 18 prompts organized by the jobs you actually do: content, email, ads, social, and analysis. Each one comes with the reason it works and the trap that makes it fail.
Why prompts matter more than tools for a solo founder
A solo founder does not need a better AI tool. The tools you already own can do most of the work. What changes the output is the instruction you give them. A vague prompt on a $200/month tool produces worse copy than a sharp prompt on the free version of Claude or ChatGPT.
Gartner estimates that by 2028, 60% of all B2B seller work will be executed using generative AI. The gap between teams that win with AI and teams that waste money on it is not the model. It is the quality of the instructions they feed it.
Salesforce's Small and Medium Business Trends Report found that 67% of marketing leaders already believe AI will help them better understand customers and boost growth. The same report shows 71% of marketers plan to use both generative and predictive AI within the next 18 months. Everyone has the tool. Almost nobody has the prompts.
The prompt is the control surface for your marketing agent. It decides the tone, the constraints, the structure, and the quality bar. Write it well and your agent produces work that sounds like you. Write it lazily and your agent produces work that sounds like everyone else.
The four-part prompt framework that actually works
Every prompt that produces usable output has four parts. Skip one and the output degrades. The four parts are role, context, task, and constraints.
Role tells the model who to be. A prompt that starts with 'you are a direct response copywriter who has written for B2B SaaS' performs differently from one that starts with 'write some copy'. The role sets the vocabulary, the priorities, and the default assumptions.
Context is everything the model needs to know but cannot guess. Your product, your audience, your offer, your past examples. Consensus calls this grounding: the more real-world context you share, the more on-brand the output becomes. If you do not give it your customer's actual words, the model invents a customer who does not exist.
Task is the single clear action. One prompt, one deliverable. 'Write a landing page headline' is a task. 'Help me with marketing' is not a task, it is a wish.
Constraints are the guardrails: length, tone, format, what to avoid. 'Under 40 words. No adjectives like revolutionary. End with a question.' The constraints turn a draft into something close enough to publish that editing takes minutes, not hours.
Use AI as a collaborator, not a generator. A generator hands you a finished product. A collaborator helps you think, draft, edit, and refine. Teams that treat AI as a generator end up with generic outputs. Teams that treat AI as a collaborator end up with sharper work and a smarter team.
That framing from IMPACT's 2026 guide is the whole game. The prompts below are written as collaboration prompts, not as 'generate me a thing' prompts. They ask the model to think, then draft, then revise. That is the difference between content that converts and content that gets skimmed and forgotten.
Content and SEO prompts that produce usable output
Content is the highest-return marketing activity for a solo founder. One blog post compounds for years. One good page can pull organic traffic for a decade. These prompts make the production loop fast without making the output generic.
Prompt 1: the outline from search intent. 'You are an SEO strategist. Here is my target keyword and the search intent behind it. Analyze the top 5 ranking pages for this keyword and give me an outline that covers every angle they cover, plus one angle none of them cover. Label each section with the question it answers.' The last instruction is the one that separates this from a lazy outline: it forces the model to think about intent, not just topic.
Prompt 2: the first draft with your voice. 'Here are three pieces of my writing. Extract my voice: sentence length, vocabulary, humor level, how I open, how I close. Then write the first draft of this post in that voice. Do not use words I would not use. Flag anything you are unsure about instead of inventing it.' Giving examples of your own writing is the fastest way to make AI sound like you.
Prompt 3: the repurpose pass. 'Turn this blog post into three LinkedIn posts, one X thread, and one newsletter intro. Keep the core claim identical in every version. Change the format, not the argument. Each version must stand alone.' One piece of content becomes five distribution assets without the model drifting off message.
Prompt 4: the SEO check before you publish. 'Here is my draft. Find every place where I make a claim without evidence, every heading that does not match the content below it, and every paragraph that could be cut without losing meaning. List them, ranked by how much they hurt the reader.' A model that audits your draft is worth more than a model that writes it, because you already know how to write. You need someone to catch the gaps.
Email and cold outreach prompts
Email is the channel a solo founder owns outright. Social platforms change the rules. Your list stays yours. Cold outreach is where prompt quality shows up fastest, because a bad prompt produces the exact generic cold email that everyone deletes.
Prompt 5: the sequence that respects the reader. 'Write a 4-email cold sequence for this persona. Email 1 is a pattern interrupt under 100 words. Email 2 shares one specific insight about their industry. Email 3 is a value-first case study. Email 4 is the breakup. Each email must have one job and one call to action. No filler sentences.' The full 5-email framework behind this prompt lives in our cold email sequences guide for solo founders.
Prompt 6: the subject line lab. 'Write 10 subject lines for the email below. Five should be curiosity-driven without being clickbait. Five should be benefit-driven without being salesy. No all caps, no emojis, no words like quick or question. Rank them by likely open rate and explain the ranking.' The explanation matters: you learn the model's reasoning instead of just picking a line.
Prompt 7: the reply-triggering close. 'Here is my cold email draft. Rewrite the final paragraph so it invites a reply instead of ending the conversation. Give me three options: one that asks a specific question about their business, one that offers a tiny piece of value with no ask, and one that references something they published recently. Explain what each option signals to the reader.'
Prompt 8: the follow-up that does not nag. 'I sent this email 5 days ago and got no reply. Write a follow-up that adds one new piece of value instead of repeating my first ask. Under 80 words. Do not mention the lack of reply. Do not apologize. End with a soft question about their priorities this quarter.' The 'do not mention the lack of reply' constraint is the part that keeps the follow-up human.
Ads and social prompts
Paid ads and social media are where generic AI output gets expensive. A bland ad wastes budget. A bland social post wastes a channel. The prompts below force the model to think about the viewer, not the format.
Prompt 9: the angle generator for ads. 'I am running ads for this product to this audience. Generate 12 distinct angles. Each angle must be a different emotional or rational hook: fear, status, speed, cost, proof, curiosity, simplicity, identity, loss, gain, ease, and social proof. For each angle, write one headline under 30 characters and one body under 90 characters. No angle may reference another angle.'
Prompt 10: the creative that matches the platform. 'Here is my core ad message. Rewrite it for LinkedIn, Instagram, and a Google search ad. LinkedIn gets a professional, insight-led version. Instagram gets a short, visual, emotional version. Google gets a tight, keyword-forward version with a clear offer. Preserve the core claim in all three. Tell me what changed and why.'
Prompt 11: the social post that starts a conversation. 'Here is my product and my audience. Write 5 social posts that each make one specific, debatable claim about my industry. No generic motivational content. No '5 tips' lists. Each post must end with a question that invites disagreement. The goal is comments from the right people, not likes from everyone.'
Prompt 12: the content calendar from your data. 'Here are my last 30 posts with their engagement numbers. Find the patterns: what topics, formats, and hooks consistently outperform. Then propose the next 10 posts that double down on the winning patterns and explain why each one is likely to work.' This turns your own post history into a repeatable content engine. Most founders never ask the model to learn from their own data.
Analysis and strategy prompts
Analysis is the most underrated use of AI for a solo founder. The model is excellent at finding patterns in messy data and at pressure-testing your reasoning. These prompts make it a thinking partner instead of a content machine.
Prompt 13: the competitor teardown. 'Here are the landing pages of my top 3 competitors. For each one, identify their positioning, their offer, their proof, and their pricing strategy. Then find the gap between what they promise and what their page actually delivers. Give me 3 opportunities they are leaving open.' The gap analysis is the part that matters. Describing competitors is easy. Finding their blind spots is valuable.
Prompt 14: the customer signal reader. 'Here are 20 customer emails, support tickets, and reviews from my product. Group them by the problem they describe. For each group, quote the exact words customers use. Then tell me which problem, if solved, would create the most marketing value. Use their language, not mine, in everything you write.' Customer words are marketing gold. This prompt extracts them without you reading 20 messages.
Prompt 15: the campaign post-mortem. 'Here is what I ran, what I spent, and what happened. Give me a brutal post-mortem: what worked, what wasted money, what I should test next, and what I should never run again. Do not soften the findings. Rank the next tests by expected impact per dollar.' Ask for brutal and you get useful. Ask for a summary and you get a highlight reel.
Prompt 16: the pricing sanity check. 'Here is my pricing and my target customer. Interview me with 10 questions that would expose whether my pricing is wrong. Ask one question at a time. After I answer all 10, give me your honest read on the pricing and what I should test.' The interview format forces you to think instead of just reading a generated opinion.
How to build a prompt library you can reuse
The biggest difference between a founder who gets value from AI and one who does not is a prompt library. A library turns your best prompts into repeatable assets. Every time you write a prompt that produces output you actually use, save it. Every time you revise a prompt because the output was weak, save the new version.
Start with a simple text file or a note app. One section per job: content, email, ads, social, analysis. Each entry has the prompt, a note on when to use it, and the best output it produced. That third field is the one most people skip and the one that makes the library useful six months later.
Version your prompts the way you version code. Prompt v1 produces generic output. Prompt v2 adds your voice examples and the output improves. Prompt v3 adds the constraint about flagging uncertainty and the output gets even better. Keep the version history. When a tool updates and your old prompt breaks, the history shows you what changed.
Steal the pattern, not just the words. When you see a prompt that works, ask why it works. Which part is the role? Which part is the constraint? Rebuild it in your own words for your own context. A prompt you understand beats a prompt you copied, because you can fix it when it breaks.
If you are new to the agent side of this, our post on AI marketing agents for solo founders covers what agents actually do and what they still cannot. The prompts above are the steering wheel. The agent post is the engine.
And if you are still choosing between the general assistants, our breakdown of AI marketing tools for solo founders ranks what to buy first by time saved per dollar, which pairs cleanly with the prompt library you build here.
Mistakes that turn good prompts into bad content
The most common mistake is prompt-and-publish. You generate a draft, skim it, and ship it. IMPACT calls this the fastest way to hurt your credibility. AI drafts are starting points, not finished work. The edit pass is where the voice comes back.
The second mistake is vague context. 'Write about my product' produces nothing useful because the model does not know your product. Every prompt needs the specifics: what it does, who it is for, what makes it different. If you are tired of repeating the context, save it in a reusable snippet and paste it into every prompt.
The third mistake is ignoring the model's assumptions. The model fills gaps with what it assumes to be true. If you do not give it your real customer language, it invents plausible-sounding language that may not match reality. Every prompt that touches facts needs the facts in the prompt.
The fourth mistake is asking for everything at once. 'Write me a full marketing strategy' is not a prompt, it is a project. Break it into single tasks: audience, positioning, channels, offer, then content. Each step gets the full four-part framework. The outputs stack into a strategy you can actually use.
The fifth mistake is never iterating. The first output is rarely the best. The fastest way to improve is to tell the model what to change: 'Make it shorter. Make it more specific. Remove the adjectives. Give me three alternatives for the opening.' Iteration is where the prompt framework pays off, because each revision sharpens the constraints.