AI prompts for content writing work best when they define the audience, objective, approved source material, tone and output format. This guide gives you 38 copy-ready templates for blog briefs, social posts, ad copy, emails and landing pages, plus checks for factual accuracy and brand safety.

Vague requests produce generic drafts because they leave the model to guess. The templates below make those decisions explicit, ask for evidence when claims are needed, and leave final publication decisions with a human editor.
This library contains 38 distinct, copy-ready prompts: 30 task-specific templates and eight reusable editing and context prompts. Treat every output as a draft to validate, not a proven performance result.
For the broader workflow behind these prompts, start with our AI content strategy guide.
For source-grounded research and content briefs, it helps to distinguish a notebook workflow from general prompt writing. Our update on the NotebookLM to Gemini Notebook change discusses the evolution of Google’s research experience.
What Makes an AI Content Writing Prompt Good?
A useful prompt supplies enough context to produce a draft a real editor can evaluate. It should name who will read the content, what the content must achieve, which information is authoritative, and how the brand should sound.
Use four fields before every request: Audience (the reader and their situation), Objective (the decision or action), Source (approved evidence), and Tone (the brand voice). If source material is missing, ask the model to identify gaps rather than invent details. For a deeper editing workflow, see our AI brand voice guide.
AI Content Prompt Example: From Inputs to Output
The example below is illustrative. It uses a fictional SaaS reporting product so that no performance results, customer claims or product capabilities are attributed to a real company.
Example input
Audience: RevOps leaders at B2B SaaS companies who currently reconcile CRM exports. Objective: Generate a short landing-page hero that encourages a qualified demo. Source: Fictional approved brief: the product connects to Salesforce, shows a daily pipeline dashboard and allows role-based views. Tone: Plainspoken, specific and helpful.
Copy-ready example prompt
Audience: RevOps leaders at B2B SaaS companies who reconcile CRM exports. Objective: Encourage qualified visitors to book a demo. Source: Fictional approved product brief. The product connects to Salesforce, provides a daily pipeline dashboard, and offers role-based views. No performance or ROI results are available. Tone: Direct, credible, and concise. Write one landing-page headline, a supporting sentence under 28 words, and a three-word CTA. Use only the details in the brief. Do not invent time savings, pipeline uplift, security certifications, or customer testimonials.
Illustrative output
Headline: One clearer view of your SaaS pipeline. Supporting copy: Bring Salesforce opportunity data into a daily pipeline dashboard, with views tailored to each team. CTA: Book a demo.
Editorial check: Confirm that the stated integration, dashboard refresh and role-based views work exactly as described before using this wording for a real product. An AI draft is not proof of functionality.
AI Content Writing Prompts by Type
Below are 30 category-specific templates across five publishing jobs. Every card includes audience, objective, source and tone inputs. Replace the bracketed fields before using a prompt, and verify the resulting content against approved materials.
AI Prompts for Blog Writing
Use these AI prompts for blog writing when you have a source-backed topic but need a stronger brief, outline, tutorial, comparison or refresh. To connect the writing to ICP problems and measurable outcomes, pair the templates with our B2B SaaS content strategy guide.
Search-Intent Content Brief
Audience: [BUYER ROLE, COMPANY STAGE, TOP PROBLEM] Objective: [INFORMATIONAL QUERY AND DESIRED READER ACTION] Source: [VERIFIED SEARCH-INTENT NOTES, PRODUCT DOCS, EXPERT INTERVIEW, AND DATA] Tone: [BRAND VOICE AND READING LEVEL] Create a blog brief for [TOPIC] targeting [PRIMARY QUERY]. State the reader's immediate question in one sentence. Propose an H1, 5-7 H2s, unique insight for each section, source evidence required, internal questions the draft should answer, and one relevant next step. Distinguish sourced facts from hypotheses. Do not invent search volumes, statistics, quotes, rankings, or client results. If the source is incomplete, list the missing evidence.
Expert Interview to Blog Outline
Audience: [READER JOB TITLE AND PRIOR KNOWLEDGE] Objective: [WHAT READER SHOULD LEARN OR DECIDE] Source: [APPROVED SME INTERVIEW NOTES OR TRANSCRIPT] Tone: [DIRECT / TECHNICAL / CONVERSATIONAL] Convert the provided expert notes about [TOPIC] into a blog outline. Identify three defensible insights, supporting examples, unresolved questions, and a logical section flow. Preserve the expert's meaning; distinguish direct quotes from paraphrases. Suggest where firsthand examples improve the article. Output headings, section bullets, and an evidence-to-claim checklist. Do not create quotes or technical details absent from the notes.
Evidence-Led Blog Introduction
Audience: [TARGET READER AND SITUATION] Objective: [WHY THE READER NEEDS THIS ARTICLE] Source: [APPROVED FACTS, PROBLEM STATEMENT, AND BRIEF] Tone: [BRAND VOICE] Write three opening paragraphs for [BLOG TOPIC], 70-110 words each. Open with the reader's problem or a specific source-supported observation. Explain what this article helps them do, without hype or invented outcomes. Avoid generic openings such as 'In today's digital world'. After the options, explain which one most closely matches the stated search intent. Flag any factual claim that needs a source.
How-To Tutorial Section
Audience: [USER ROLE AND TECHNICAL LEVEL] Objective: [TASK THE READER MUST COMPLETE] Source: [CURRENT VERIFIED PRODUCT DOCUMENTATION OR APPROVED SCREENSHOTS] Tone: [CALM, PRECISE INSTRUCTIONAL VOICE] Draft one practical how-to section on [TASK]. Include prerequisites, numbered steps using only documented interface labels, expected result, common error states, and a final verification step. Use 'check in your account' if interface options depend on plan or permissions. Do not guess menu names, dates, default settings, or feature availability. Mark missing screenshots or unresolved steps for editorial review.
Comparison Article Framework
Audience: [EVALUATOR ROLE AND USE CASE] Objective: [DECISION BETWEEN OPTIONS] Source: [VERIFIED FEATURE LISTS AND FIRST-PARTY PRICING OR POLICY NOTES] Tone: [NEUTRAL, PRACTICAL] Build a comparison outline for [OPTION A] versus [OPTION B] on [TOPIC]. Compare best-fit team, requirements, useful capabilities, limitations, total-workflow considerations, and questions to ask before purchase. Separate confirmed facts from editorial judgment. Produce a concise decision matrix and a 'choose this when' paragraph for each option. Do not fabricate feature, price, customer, or performance claims.
Existing Blog Refresh and Claim Audit
Audience: [CURRENT ARTICLE'S PRIMARY SEARCH READER] Objective: [IMPROVE ACCURACY WITHOUT LOSING EXISTING INTENT] Source: [CURRENT ARTICLE, APPROVED CHANGE LOG, RECENT SEARCH DATA, AND VERIFIED REFERENCES] Tone: [PRESERVE ESTABLISHED BRAND VOICE] Review [PASTE EXISTING ARTICLE] for stale claims, duplicate sections, thin explanations, mismatched intent, broken logic, and potential keyword cannibalization. Return a Keep / Replace / Remove / Add plan at paragraph level. Preserve verified examples, current internal links, and working widgets. Do not rewrite established ranking sections without supporting evidence. List every claim and URL requiring human verification before publication.
Fact-check and brand-safety review: Confirm search intent and query ownership, verify statistics against original sources, check quotes and screenshots with subject-matter experts, and remove unsupported ranking or traffic predictions.
AI Prompts for Social Media Content
Social prompts should adapt one approved idea to the format and attention pattern of each channel. For visual storytelling, our existing Instagram carousels resource covers visual-prompt techniques.
LinkedIn Insight Post
Audience: [BUYER PERSONA AND INDUSTRY] Objective: [START AN INFORMED CONVERSATION ABOUT ONE IDEA] Source: [APPROVED BLOG EXCERPT, INTERVIEW, OR FIRSTHAND OBSERVATION] Tone: [BRAND OR FOUNDER VOICE] Draft one LinkedIn post of 130-180 words about [TOPIC]. Lead with a source-supported tension, give one concrete example, and end with one genuine discussion question. Keep paragraphs easy to scan. Avoid engagement bait, fake personal experience, made-up statistics, and generic motivational claims. Provide two alternate opening lines and note anything that must be fact-checked.
Founder Point-of-View Post
Audience: [RELEVANT DECISION MAKERS] Objective: [EXPLAIN A DEFENSIBLE PERSPECTIVE] Source: [FOUNDER NOTES, APPROVED STORY, AND FIRSTHAND OBSERVATIONS] Tone: [PERSONAL, DIRECT, CREDIBLE] Write a founder-style post about [ISSUE]. Structure it as an observed problem, what we learned, what we would do differently, and a practical takeaway. Use first person only for experiences explicitly recorded in the source. Produce one 160-word draft and three headline hooks. Do not invent a founder anecdote, customer quote, or commercial result.
X Thread from a Research Note
Audience: [COMMUNITY AND EXPERTISE LEVEL] Objective: [TEACH ONE WORKFLOW OR INSIGHT] Source: [APPROVED STUDY SUMMARY OR COMPANY RESEARCH NOTE] Tone: [CONCISE, NON-SENSATIONAL] Turn [SOURCE NOTE] into a thread of six posts for X. Each post should convey one useful idea and work without exaggerated claims. Start with the main takeaway, explain context and limits, and close with an application question. Label data sources in the draft notes rather than inventing citations. Check individual post lengths against the applicable account limits before scheduling.
Instagram Carousel Script
Audience: [AUDIENCE SEGMENT AND PAIN POINT] Objective: [TEACH ONE PROCESS VISUALLY] Source: [APPROVED HOW-TO CONTENT] Tone: [CLEAR, FRIENDLY, BRAND-SAFE] Create a seven-slide Instagram carousel about [TOPIC]. For each slide, provide one short on-image line, optional supporting line, and a visual direction that does not require licensed third-party imagery. Use slide 1 for a useful promise, slides 2-6 for distinct steps, and slide 7 for a helpful next step. Add a 90-130 word caption. Do not claim outcomes or use unlicensed testimonials.
Short-Form Video Hook and Script
Audience: [VIEWER PROFILE] Objective: [EXPLAIN ONE BENEFIT OR TECHNIQUE] Source: [APPROVED PRODUCT DEMO OR EDITORIAL BRIEF] Tone: [NATURAL SPOKEN BRAND VOICE] Draft a roughly 30-second vertical-video script for [TOPIC], with a first-three-second hook, three short beats, on-screen text, a visual shot idea for each beat, and one accurate CTA. Use only demonstrated features and accessible visual elements. Do not promise virality or write an unsupported before-and-after story. Include a note on subtitles and brand review.
Webinar to Social Distribution Plan
Audience: [FOLLOWERS ACROSS LINKEDIN, X, AND INSTAGRAM] Objective: [DRIVE USEFUL ENGAGEMENT WITH AN APPROVED WEBINAR] Source: [APPROVED WEBINAR TRANSCRIPT AND SUMMARY] Tone: [BRAND VOICE, ADAPTED PER CHANNEL] Extract one strong insight from [WEBINAR EXCERPT] and adapt it into a LinkedIn text post, a four-post X thread, and a five-slide Instagram concept. Do not duplicate the same words across formats. Name any source references and distinguish guest opinions from verified facts. Add one channel-specific CTA to [VERIFIED DESTINATION]. Do not invent speaker quotes or expand claims beyond the webinar.
Fact-check and brand-safety review: Confirm guest attribution, reuse and image rights, claims in slides or videos, platform formatting, and whether any firsthand story is genuinely documented. Adapt rather than duplicating copy across channels.
AI Prompts for Ad Copy
Use these prompts to generate testable creative, not assured performance. Match the copy to the approved offer and landing page. For implementation support across channels, explore OneMetrik’s Google Ads, Meta and LinkedIn services.
Google Search Responsive Ad Assets
Audience: [SEARCHER INTENT AND CUSTOMER PROFILE] Objective: [QUALIFIED CLICK OR DEMO ACTION] Source: [APPROVED OFFER, LANDING PAGE, PRODUCT FEATURES, AND KEYWORDS] Tone: [CLEAR, SPECIFIC, COMPLIANT] Draft 12 distinct Google Search responsive ad headlines and four descriptions for [OFFER]. Use only landing-page-supported benefits. Target no more than 30 characters for each headline and 90 for each description; report character counts and flag any over-limit lines for manual review. Vary message angles without inventing pricing, guarantees, or urgency. Check final assets against current platform rules and the destination page before launch.
Meta Feed Ad Concept
Audience: [TARGET SEGMENT AND PROBLEM] Objective: [CLICK TO APPROVED OFFER] Source: [APPROVED PRODUCT BRIEF AND LANDING PAGE] Tone: [HUMAN, BRAND-SAFE, SPECIFIC] Create three Meta feed ad concepts for [PRODUCT OR SERVICE]. For each, provide a 60-90 word primary-text option, a short headline, a visual brief, and the single offer-backed benefit. Use inclusive language and avoid assuming sensitive personal characteristics about viewers. Do not fabricate testimonials, discounts, product capabilities, or results. List claims that require legal or policy review.
LinkedIn Sponsored Content
Audience: [BUSINESS ROLE, SENIORITY, AND WORK PROBLEM] Objective: [LEAD MAGNET DOWNLOAD OR QUALIFIED DEMO] Source: [APPROVED REPORT OR PRODUCT ONE-PAGER] Tone: [PRACTICAL, CREDIBLE, B2B] Write three LinkedIn sponsored-content variations for [ASSET OR OFFER]. Make each variation use a different evidence-backed angle: problem, framework, or example. Provide an opening line, primary body copy, CTA label, and suggested supporting creative. Avoid invented company proof, misleading personalization, and performance guarantees. Confirm that each claim appears in the linked landing page.
Retargeting Reminder Ad
Audience: [CONSENTED AUDIENCE DEFINED BY CAMPAIGN RULES] Objective: [RE-EXPLAIN VALUE AND RETURN TO LANDING PAGE] Source: [APPROVED OFFER, PRODUCT FAQ, AND POLICY REVIEW] Tone: [HELPFUL, LOW-PRESSURE] Draft three retargeting ad messages for [OFFER], focusing on one practical benefit or unanswered question. Do not mention or imply that the brand tracked an individual's behavior, job search, health, financial circumstances, or browsing history. Avoid creepy 'we saw you' language. Output copy, corresponding creative idea, CTA, and a brief privacy-and-accuracy check.
15-Second Video Ad Script
Audience: [PLATFORM VIEWER PROFILE] Objective: [UNDERSTAND OFFER AND TAKE ONE ACTION] Source: [APPROVED DEMO FOOTAGE, PRODUCT CLAIMS, AND OFFER DETAILS] Tone: [DIRECT, SPOKEN, NON-HYPERBOLIC] Write two 15-second video-ad script options for [OFFER]. For each, show a hook, visual actions, voiceover, on-screen labels, and final CTA with approximate timestamps. Use only footage and product claims we can substantiate. Avoid unlicensed imagery or music, artificial testimonials, and unverifiable 'best' claims. Flag what needs platform and legal review.
Controlled Ad Copy Experiment
Audience: [CAMPAIGN ICP] Objective: [TEST ONE CLEAR HYPOTHESIS] Source: [CURRENT BASELINE AD, APPROVED OFFER, AND EXPERIMENT BRIEF] Tone: [MATCH EXISTING BRAND VOICE] Create an A/B test plan for [AD COPY]. Change only the value-proposition angle while keeping the offer, CTA, and audience assumptions consistent. Provide control copy, one challenger, why the challenger could resonate, the primary quality metric [QUALIFIED LEADS / PIPELINE], and a review checklist. Do not predict a percentage lift or invent results. Treat test performance as unknown until measured.
Fact-check and brand-safety review: Confirm every offer against its landing page, check character counts in the ad platform, review creative rights and policies, and avoid sensitive-attribute targeting language and guaranteed results.
AI Prompts for Email Marketing
These email prompts cover permission-based welcome, nurture, newsletter, event and business outreach copy. The goal is useful communication that accurately reflects your offer and respects the recipient.
Welcome Email after Signup
Audience: [OPTED-IN NEW USER OR SUBSCRIBER] Objective: [HELP USER COMPLETE ONE FIRST STEP] Source: [APPROVED ONBOARDING FLOW, PRODUCT DOCS, AND SUPPORT LINK] Tone: [WARM, HELPFUL, PLAIN LANGUAGE] Draft a welcome email for [PRODUCT OR NEWSLETTER]. Include two subject options, preview text, a short body, one immediate action, and a sign-off. Explain what recipients can expect next only if the workflow is confirmed. Avoid false sender familiarity, unsupported product capabilities, and additional promotional promises. Check all links and personalization fields.
Value-Led Newsletter Issue
Audience: [EXISTING OPTED-IN SUBSCRIBERS] Objective: [HELP READERS APPLY ONE INSIGHT] Source: [APPROVED BLOG ARTICLE, ORIGINAL DATA, OR RECORDED EXPERT NOTE] Tone: [EDITORIAL, CONCISE, INFORMED] Write a 200-250 word newsletter about [TOPIC]. Lead with one valuable conclusion, share an approved example, explain a practical application, and link to [VERIFIED RESOURCE] with a single CTA. Offer three subject lines that accurately reflect the content. Do not invent figures, research citations, partner quotes, or urgency.
Three-Email Educational Nurture
Audience: [OPTED-IN B2B LEAD AND BUYING STAGE] Objective: [BUILD UNDERSTANDING BEFORE A DEMO INVITATION] Source: [APPROVED CASE DETAILS, PRODUCT BRIEF, AND RESOURCE LINKS] Tone: [RESPECTFUL, BUYER-CENTRIC] Create three connected nurture emails for [BUYER PROBLEM]: first explain the problem, second give a useful evaluation framework, third introduce a relevant next step. Give each email a subject, preview, 100-150 word body, and one CTA. Do not fabricate proof or claim a specific result. Keep the series useful even for readers who are not ready to buy.
Permission-Based Re-Engagement
Audience: [CONTACTS WITH VALID CONSENT AND DEFINED SEGMENT] Objective: [OFFER A RELEVANT CHOICE OR PREFERENCE UPDATE] Source: [APPROVED CUSTOMER COMMUNICATION POLICY AND NEW RESOURCE] Tone: [POLITE, LOW-PRESSURE] Write a re-engagement email for recipients who have not interacted with [NEWSLETTER OR PRODUCT] recently. Offer one genuinely useful update and a transparent choice to continue, adjust preferences, or unsubscribe. Do not use guilt, fake deadlines, surveillance language, or imply knowledge we do not have. Provide two subject lines and a concise body.
Research-Based B2B Introduction Email
Audience: [BUSINESS ROLE AND COMPANY CONTEXT] Objective: [REQUEST A SHORT, RELEVANT CONVERSATION] Source: [PUBLIC, VERIFIED COMPANY NOTE AND APPROVED SERVICE DETAILS] Tone: [PROFESSIONAL, PERSONAL BUT NOT INTRUSIVE] Write a 90-120 word first-contact email for [RECIPIENT ROLE]. Use only the provided verified business context to explain one relevant reason for contacting them. Propose a low-pressure next step. Provide a subject line and one follow-up. If recipient or company detail is missing, use [RESEARCH NEEDED] instead of invented personalization. Ensure the final send respects applicable outreach and consent requirements.
Webinar Invitation Email
Audience: [SUBSCRIBERS INTERESTED IN TOPIC] Objective: [REGISTER FOR AN ACTUAL EVENT] Source: [CONFIRMED EVENT TITLE, DATE, TIME ZONE, SPEAKERS, AND REGISTRATION URL] Tone: [PRACTICAL AND INVITING] Draft an invitation email for [EVENT]. Include one specific learning outcome, three agenda points supported by the event brief, presenter names only when confirmed, date and time zone, and one registration CTA. Provide two subject lines and preview text. If information is missing, use a clearly labeled placeholder rather than inventing a guest speaker, scarcity, or deadline.
Fact-check and brand-safety review: Check permission and suppression rules, accurate sender identity, working preference/unsubscribe links, confirmed event details and CRM merge fields. Do not insert fictional personalization.
AI Prompts for Landing Page Copy
Landing-page prompts should map one offer to one visitor problem and one action. Supply actual product details, approved proof and form behavior. Do not invent testimonials or conversion-rate promises.
SaaS Landing Page Hero
Audience: [TARGET VISITOR AND AWARENESS STAGE] Objective: [REQUEST DEMO OR START TRIAL] Source: [APPROVED PRODUCT POSITIONING, ACTUAL FEATURES, AND OFFER] Tone: [DIRECT, BENEFIT-LED, BRAND VOICE] Draft five landing-page hero options for [PRODUCT]. Each must contain a clear headline, one supporting sentence, one primary CTA, and a short reason to believe based only on supplied evidence. Reflect the visitor's problem and avoid invented customer outcomes. Present a short note explaining the primary distinction between the options.
Feature-to-Benefit Section
Audience: [BUYER ROLE AND CURRENT WORKFLOW] Objective: [MAKE PRODUCT VALUE CONCRETE] Source: [VERIFIED FEATURE DOCUMENTATION AND CUSTOMER JOBS TO BE DONE] Tone: [PLAIN, SPECIFIC, NON-SALESY] Turn [THREE APPROVED FEATURES] into a three-item benefits section. For each, include a user problem, what the feature actually does, the practical consequence, and the caveat or setup prerequisite where relevant. Do not imply an integration, automation, service level, or quantified ROI that the source does not support.
Landing Page Objection Answers
Audience: [EVALUATOR AT BUYING STAGE] Objective: [RESOLVE ACTUAL SALES OBJECTIONS] Source: [APPROVED SALES QUESTIONS, SECURITY NOTES, PRICING POLICY, AND PRODUCT DOCS] Tone: [CALM, TRANSPARENT, PRECISE] Write four short answers to these documented objections: [PASTE OBJECTIONS]. Each answer should state the relevant fact first, then explain any limitations or conditions, and close with a useful next step where applicable. If the answer depends on legal, implementation, or security review, label that dependency. Do not make certification, guarantee, or price claims not present in source materials.
Lead Magnet Signup Page
Audience: [READER SEEKING SPECIFIC RESOURCE] Objective: [DOWNLOAD AN APPROVED GUIDE] Source: [FINAL RESOURCE CONTENT AND ACTUAL FORM FIELDS] Tone: [CLEAR, HELPFUL, PRIVACY-CONSCIOUS] Draft a landing page for [RESOURCE TITLE]. Include the one-line resource promise, three accurate takeaways, suitable-audience line, form intro, and one download CTA. Describe only what the resource really contains. Do not guarantee outcomes, invent scarcity, or suggest collecting personal data that the existing form does not require. Include a note to verify privacy copy and form behavior.
Book-a-Demo Page
Audience: [B2B EVALUATOR AND BUYING COMMITTEE] Objective: [BOOK A QUALIFIED DEMO] Source: [APPROVED DEMO AGENDA, PRODUCT BRIEF, AND SALES PROCESS] Tone: [CONFIDENT, LOW-PRESSURE] Write a demo-request section with a specific headline, two-sentence value proposition, a three-step explanation of what happens after booking, and an accurate CTA. Add two objection-handling lines based on [APPROVED SALES NOTES]. Do not promise a meeting duration, implementation timeline, case-study result, or follow-up process unless it is documented.
Pricing and Proof Section without Fabrication
Audience: [PROCUREMENT AND BUSINESS EVALUATORS] Objective: [HELP THEM ASSESS FIT] Source: [CURRENT PRICING POLICY, PLAN TABLE, AND APPROVED TESTIMONIALS] Tone: [TRANSPARENT, FACTUAL] Write a landing-page comparison of [OPTIONS OR PLANS]. List verified inclusions, meaningful exclusions, who each fits, and next steps. Use testimonial wording only if the exact approved quote and consent are provided; otherwise omit testimonials entirely. Do not create customer logos, ratings, discounts, guarantees, or price estimates. State any information that needs sales confirmation.
Fact-check and brand-safety review: Validate product features, prices, security and integration claims, testimonials and consent, accessibility, form behavior, and CTA destinations. Do not add proof placeholders to the live page.
Universal AI Prompts for Content Writing
These eight reusable modifiers help set context, build briefs, review drafts and improve specificity. Combine them with the 30 category templates above, and provide real source material rather than asking a model to invent proof.
Voice and Tone Setup
Audience: [WHO WILL READ THE OUTPUT] Objective: [WHAT THIS PROMPT SHOULD ACHIEVE] Source: [APPROVED SOURCE MATERIAL OR NONE] Tone: [BRAND VOICE / STYLE] I'm writing content for [BRAND/COMPANY]. Our voice is: [DESCRIBE VOICE - e.g., "direct, slightly irreverent, expert but not condescending"] We sound like: [COMPARISON - e.g., "a smart colleague explaining something over coffee"] We never sound like: [ANTI-COMPARISON - e.g., "a corporate whitepaper or LinkedIn influencer"] Words/phrases we avoid: [LIST BANNED PHRASES] Use this voice for all content I ask you to write in this conversation. Use verified facts only. Do not invent numbers, quotations, product capabilities, or customer outcomes. If supporting evidence is missing, flag it for review.
Audience Context Block
Audience: [WHO WILL READ THE OUTPUT] Objective: [WHAT THIS PROMPT SHOULD ACHIEVE] Source: [APPROVED SOURCE MATERIAL OR NONE] Tone: [BRAND VOICE / STYLE] My target audience is [JOB TITLE/ROLE] at [COMPANY TYPE]. Their situation: [WHAT THEY'RE DEALING WITH] Their goal: [WHAT THEY WANT TO ACHIEVE] Their main obstacle: [WHAT'S STOPPING THEM] Their sophistication level: [BEGINNER/INTERMEDIATE/EXPERT] They already know: [WHAT YOU CAN ASSUME] They don't know: [WHAT YOU NEED TO EXPLAIN] Use verified facts only. Do not invent numbers, quotations, product capabilities, or customer outcomes. If supporting evidence is missing, flag it for review.
Content Brief Expander
Audience: [WHO WILL READ THE OUTPUT] Objective: [WHAT THIS PROMPT SHOULD ACHIEVE] Source: [APPROVED SOURCE MATERIAL OR NONE] Tone: [BRAND VOICE / STYLE] Expand this brief into a detailed content outline: Topic: [TOPIC] Format: [BLOG POST, EMAIL, LANDING PAGE, ETC.] Audience: [WHO] Goal: [WHAT THE CONTENT SHOULD ACCOMPLISH] Key points to cover: [LIST] Word count: [TARGET] Include: - Suggested headline options - Section breakdown with key points for each - Places where I need data, examples, or proof - Potential objections to address - CTA recommendation Use verified facts only. Do not invent numbers, quotations, product capabilities, or customer outcomes. If supporting evidence is missing, flag it for review.
First Draft Review
Audience: [WHO WILL READ THE OUTPUT] Objective: [WHAT THIS PROMPT SHOULD ACHIEVE] Source: [APPROVED SOURCE MATERIAL OR NONE] Tone: [BRAND VOICE / STYLE] Review this draft: [PASTE DRAFT] Check for: 1. Does it hook in the first two sentences? 2. Is every paragraph earning its place? 3. Are claims specific or vague? 4. Does it sound like [BRAND VOICE] or generic AI? 5. Is there a clear next step for the reader? Flag weak sections and suggest specific improvements. Use verified facts only. Do not invent numbers, quotations, product capabilities, or customer outcomes. If supporting evidence is missing, flag it for review.
Rewrite for Different Audience
Audience: [WHO WILL READ THE OUTPUT] Objective: [WHAT THIS PROMPT SHOULD ACHIEVE] Source: [APPROVED SOURCE MATERIAL OR NONE] Tone: [BRAND VOICE / STYLE] Here's content written for [ORIGINAL AUDIENCE]: [PASTE CONTENT] Rewrite it for [NEW AUDIENCE]. Adjust: - Terminology and jargon level - Examples and references - Pain points emphasized - Sophistication of explanations - CTA and next steps Keep the core message. Change how it's delivered. Use verified facts only. Do not invent numbers, quotations, product capabilities, or customer outcomes. If supporting evidence is missing, flag it for review.
Trim Without Losing Value
Audience: [WHO WILL READ THE OUTPUT]
Objective: [WHAT THIS PROMPT SHOULD ACHIEVE]
Source: [APPROVED SOURCE MATERIAL OR NONE]
Tone: [BRAND VOICE / STYLE]
This content is [CURRENT WORD COUNT] words. Cut it to [TARGET WORD COUNT] without losing the key points.
[PASTE CONTENT]
Priorities:
1. Keep all verified numbers and approved examples
2. Keep the main argument intact
3. Cut redundancy and filler first
4. Combine sentences where possible
5. Remove hedge words ("basically," "essentially," "in order to")
Use verified facts only. Do not invent numbers, quotations, product capabilities, or customer outcomes. If supporting evidence is missing, flag it for review.
Add Specificity
Audience: [WHO WILL READ THE OUTPUT] Objective: [WHAT THIS PROMPT SHOULD ACHIEVE] Source: [APPROVED SOURCE MATERIAL OR NONE] Tone: [BRAND VOICE / STYLE] This content is too generic. Make it specific: [PASTE CONTENT] Add: - Specific numbers where you've used vague language - A concrete example for each main point - Names of tools, companies, or methods where relevant - Timeframes and quantities instead of "fast" or "many" If the source does not provide a number or example, write [EVIDENCE NEEDED] rather than inventing a plausible detail. Use verified facts only. Do not invent numbers, quotations, product capabilities, or customer outcomes. If supporting evidence is missing, flag it for review.
Competitor Differentiation Angle
Audience: [WHO WILL READ THE OUTPUT] Objective: [WHAT THIS PROMPT SHOULD ACHIEVE] Source: [APPROVED SOURCE MATERIAL OR NONE] Tone: [BRAND VOICE / STYLE] I'm writing about [TOPIC]. Most content on this topic says [COMMON ADVICE]. My angle: [YOUR DIFFERENT PERSPECTIVE] Why I believe this: [YOUR REASONING OR EVIDENCE] Help me write content that: - Acknowledges the common advice - Explains why it's incomplete or wrong - Presents my alternative with specifics - Doesn't sound contrarian just for clicks Format: [BLOG POST, LINKEDIN POST, ETC.] Use verified facts only. Do not invent numbers, quotations, product capabilities, or customer outcomes. If supporting evidence is missing, flag it for review.
How to Use AI Writing Prompts Effectively
Start with source-backed context. Supply approved facts, audience detail and real voice examples before requesting an outline or draft. A prompt cannot substitute for material you do not have.
Specify format and constraints. State intended length, output structure, brand vocabulary, exclusions and the review criteria. Treat model-generated counts as suggestions until checked.
Draft in stages. Approve the outline, work section by section and conduct separate passes for evidence, usefulness and voice. A blog post in one prompt is rarely a sufficient expert-led review workflow.
Use human editorial review. Check every substantive fact, quote, citation, offer and policy-sensitive detail. Editing effort depends on the task and evidence, so avoid universal time-saving claims.
Save approved templates. Document the intended audience, goal, source notes, review owner and version of each prompt so your team can reuse the workflow safely.
When adapting a strong asset into multiple channels, write a native brief for each rather than merely paraphrasing. Our guide to iterative content provides a broader content repurposing framework.
Common Mistakes to Avoid
Skipping subject-matter review: Technical, legal and product claims need an expert when the editor cannot verify them independently.
Accepting the first draft: Compare output with the approved brief, fix missing evidence and remove repetition before polishing style.
Leaving out buyer context: State the reader task, knowledge level, buying stage and desired next decision.
Confusing tone with expertise: Supply authentic brand examples, then review perspective, terminology and empathy by hand.
Publishing without QA: Check privacy, permissions, factual accuracy, copyrights, internal links, accessibility and desktop/mobile formatting.
AI Writing Fact-Check and Brand-Safety Checklist
Before publishing AI-assisted content, confirm that every factual claim has an approved source; product features, offers, prices and dates match documentation; any customer story or testimonial is authorized; inputs meet confidentiality and privacy requirements; and images, music, quotations and reused language have appropriate rights. Assign a human editor to approve the final content.
For ads, review platform rules and destination consistency. For email, verify opt-in, sender identity, personalisation and unsubscribe behavior. For landing pages, test form behaviour and CTA links. For blogs and tutorials, review source evidence, search intent, outdated statements and working internal links.
Best AI Tools for Content Writing Prompts
ChatGPT, Claude and Gemini can all assist with drafting, editing, synthesis and idea development. Evaluate each based on your workflow, permitted data handling and account features. Do not assume any model provides verified facts or superior writing for every task.
- ChatGPT: Use structured briefs to draft, iterate and review sections. Features and available models change over time and vary by account.
- Claude: Use supplied documents and notes to organize, summarize and revise longer drafts. Compare important details with the original source.
- Gemini: Use for ideation, drafting, summarization and, where supported, connected research workflows. Independently verify factual claims, citations and current feature details.
For teams, keep a shared, versioned prompt library with approved examples, source requirements, review owners and known limitations. Choosing a reliable process matters more than declaring one tool the universal winner.
More Prompt Resources
Looking for prompts for specific AI tools? We’ve got dedicated libraries:
Best ChatGPT Prompts
Browse ChatGPT prompt examples for writing, planning and analysis.
Best Claude Prompts
Explore Claude prompt examples for drafting and editing.
Best Gemini Prompts
Explore Gemini prompt examples for research and drafting.
Full Prompt Library
Browse prompt examples by tool and content use case.
AI prompts are starting points, not published assets. Useful content still depends on evidence selection, expert judgment, responsible editing and measurement of real buyer outcomes.
Frequently Asked Questions
What’s the best AI tool for content writing prompts?
ChatGPT handles complex prompt structures best, while Claude produces more natural first drafts. For research-heavy content, use Gemini since it accesses real-time web data. Most teams use ChatGPT for iteration and Claude for initial drafts.
Do AI writing prompts work for technical content?
Yes, but only if you can fact-check the output. Use AI for structure and first drafts, then verify all technical details yourself. Never publish AI-generated technical content without expert review.
How long should an AI content writing prompt be?
Effective prompts are typically 50-200 words. Include audience context, output format, word count, and 2-3 specific requirements. Longer prompts with examples work better than short vague ones.
For the bigger picture on how AI fits into content marketing, read our guide on AI content marketing.
Want a consistent workflow for briefs, approved sources, drafting and editorial QA? OneMetrik can help connect prompt templates to your content marketing process and B2B business goals.
Ready to Build a More Reliable AI Content Workflow?
Build an AI-assisted content workflow grounded in your audience, approved sources, brand voice and human review. OneMetrik can help you connect production to practical B2B growth goals.