Home Blog AI Prompting for Marketers
AI

AI Prompting For Marketers.

Most marketers using AI are getting back exactly what they asked for: generic, hedged, average. The problem isn't the model — it's the prompts. Here's the prompting framework I use across ChatGPT, Claude, and Perplexity to produce work that's actually publishable.

John Michael Lamigo
John Michael Lamigo
WordPress Strategist · Funnel Architect
Published Feb 07, 2026 · 11 min read
Table of Contents

Why Most Marketers Get Bad AI Output

The single biggest reason marketers say "AI content sounds robotic" is that they're prompting like they're using Google Search. "Write a blog post about email marketing" produces a generic blog post about email marketing. The AI is doing exactly what was asked — the ask was just bad.

The skill of prompting isn't a trick. It's information density. The more relevant context, constraints, and examples you give the model, the more specific and useful the output gets.

If your prompt is shorter than the output you want, your prompt is probably too short. Real working prompts often run 200-500 words to produce 1500-word outputs. The ratio of input quality to output quality is roughly linear.

The 5-Part Prompt Framework

Every prompt I write for marketing work follows the same structure. Skip any part and quality drops.

Part 1: Role + Context

Tell the AI who it is and what situation it's operating in. Don't say "act as an SEO expert." Say:

"You're a senior SEO strategist with 10+ years of experience optimizing B2B SaaS sites. You write for marketing leaders at companies doing $5–50M in revenue. They've already heard the basics — they need depth and contrarian insight."

Part 2: The Specific Task

Be precise about what you want done. Not "write a blog post." Instead:

"Write a 1,200-word blog post titled 'Why Most B2B SaaS Content Strategies Stall at $1M ARR.' The post should walk through 4 specific reasons content stalls, with one concrete fix per reason."

Part 3: Constraints and Format

Define what the output must look like:

Part 4: Voice and Examples

Give the model a writing sample to emulate. Paste 200–400 words of actual content you've written or admire. Tell it: "Match this voice." Models are exceptional at matching style when given a reference.

Part 5: Anti-Patterns

Tell the AI what you don't want. Explicit "don't" instructions matter more than positive ones with current models:

⚡ Need This Implemented?
AI-Augmented Content
Workflows. Set Up Right.
I help marketing teams set up AI workflows that produce publishable output — prompt libraries, content review systems, and integration into WordPress. Faster output, no AI slop.
Talk About AI Workflows →

10 Real Prompts for Marketing Tasks

Prompt 1: Blog Post Outline

"You're a senior content strategist who writes for [audience]. I'm targeting the keyword '[keyword]' with primary search intent of [informational/commercial/etc]. Generate a detailed outline for a 2,000-word blog post that ranks for this keyword. Include: H1, H2 sections (5–7), 2–4 H3s under each H2, and a one-line note on what each section should contain. Avoid generic subheadings — every section must have a clear, specific angle."

Prompt 2: Email Subject Line Generation

"Generate 25 subject lines for an email promoting [offer]. The audience is [audience]. Constraints: under 50 characters, no clickbait, no emoji, no all-caps. Mix curiosity, specificity, and benefit-driven angles. Don't repeat structural patterns more than 3 times. After listing them, rank your top 5 with a one-line rationale."

Prompt 3: Customer Persona Refinement

"Below is a customer interview transcript. Extract the customer's exact language for: (1) the problem they're solving, (2) the alternatives they considered, (3) what made them choose us, (4) their biggest fear before buying, (5) the words they'd use to describe the product to a peer. Use direct quotes wherever possible. [Paste transcript]"

Prompt 4: Landing Page Copy Variants

"Below is a draft of my landing page hero section. Generate 5 alternative versions, each emphasizing a different angle: (1) outcome-focused, (2) problem-focused, (3) authority-focused, (4) speed/simplicity-focused, (5) contrarian. Keep each version under 30 words for the headline and 50 for the subhead. [Paste draft]"

Prompt 5: SEO Title Tag Optimization

"Below are 10 of my blog post titles. For each: (1) rate the title 1–10 for click-through potential, (2) suggest 2 stronger alternatives under 60 characters, (3) explain in one line why your alternative beats the original. Focus on specificity, numbers, and curiosity gaps over keyword stuffing. [Paste titles]"

Prompt 6: Competitor Content Gap Analysis

"Below is the content from my top competitor's blog post on [topic] and the URL of the post. Identify: (1) what they cover well, (2) what they miss or cover poorly, (3) what fresh angles I could take that they haven't, (4) specific data, examples, or sources I could include to make my version more authoritative. [Paste content]"

Prompt 7: Ad Copy Generation

"Generate 8 Meta ad variants for [product/service]. Format: 25-character headline, 90-character primary text, 30-character description. Audience: [audience]. The angle should be [angle]. Mix emotional, logical, social proof, and direct response styles — 2 of each. Don't use 'unlock,' 'leverage,' 'game-changer,' or any superlative without justification."

Prompt 8: Re-Writing for Voice

"Below is a piece of content. Rewrite it to match the voice of the writing sample I'm pasting after. Keep the structure and key points, but adjust word choice, sentence rhythm, and specific phrases to match the sample. Read both before rewriting and identify the 3 most distinctive voice characteristics of the sample first. [Content][Sample]"

Prompt 9: Structured Lead Magnet Outline

"I'm building a lead magnet for [audience] focused on [outcome]. Generate a detailed outline for a 10-page PDF that delivers genuinely actionable value (not surface-level tips). Include: title, subtitle, 5–7 main sections, page-by-page breakdown of what's covered, and one specific exercise or template per section the reader can use. The PDF should make the reader feel they got more than expected."

Prompt 10: Audit + Critique

"Below is a draft I wrote. Critique it the way a strict editor would. Point out: (1) any sentences that say nothing, (2) clichés or filler phrases, (3) places where I'm hedging when I should be direct, (4) sections that could be cut entirely. Then rewrite the weakest 200 words. Don't be polite — be useful."

My Weekly AI Workflow

Monday: Research & Outlining

Use Perplexity for fresh research with citations. Use Claude to outline 3–5 pieces of content for the week based on the research and target keywords.

Tuesday–Wednesday: Drafting

Each piece gets a structured prompt with my voice sample and the outline. AI produces a 70% draft. I edit aggressively — usually rewriting 30–40% by hand to add specifics, examples, and personal voice.

Thursday: Critique & Optimization

Run the editor critique prompt (#10 above) on each draft. Implement the strongest critiques. Run SEO optimization prompts on title tags, meta descriptions, and H2s.

Friday: Publish + Distribute

Publish in WordPress. Generate social variants for LinkedIn, X, and email using the distribution prompts. Schedule.

This workflow lets me produce 3–5 high-quality long-form pieces per week — versus 1 per week without AI assistance. The output is genuinely better, not just faster, because the AI handles the structural work and I focus on the voice and specifics that AI can't replicate.

Which AI Tools for Which Task

TaskBest ToolWhy
Long-form content draftsClaudeBest at voice matching, longer context, less generic
Quick research with citationsPerplexityReal-time web access with sources
Email + ad copy variantsChatGPT or ClaudeBoth excellent for short-form
SEO researchPerplexity + AhrefsCombination beats either alone
Image generationMidjourney or DALL-EFor hero images, illustrations
Voice/audioElevenLabsBest voice quality currently
Code/automationClaude or ChatGPTEither works; Claude slightly better for explanation

Mistakes to Avoid

  1. Treating AI output as final. Always edit. Always add personal specifics. Pure AI text fails the smell test for any informed reader.
  2. Using the same tool for everything. Different models excel at different tasks. Match tool to job.
  3. Vague prompts. "Make it better" is not a prompt. Specific criticism with examples is.
  4. Ignoring AI's strengths. AI is excellent at: outlining, brainstorming, editing structure, generating variants, and matching voice when given samples. Use those superpowers.
  5. Pretending AI didn't help. The most useful workflow is "AI accelerates my work" not "AI replaces my work." Be honest with yourself about how much editing each piece needs.

The marketers winning with AI in 2026 aren't the ones using AI more — they're the ones using it more skillfully. The skill is prompting, not pressing buttons.

Frequently Asked Questions

Will Google penalize AI-written content?
Google's stance is that the focus is on quality, not how content was made. Helpful, original, expert content ranks regardless of how it was drafted. Generic, low-effort AI content gets filtered — but so does generic, low-effort human content. The penalty isn't for using AI; it's for publishing slop.
Which AI is best for marketing in 2026?
Claude (especially the latest Claude Opus models) generally produces the best long-form content with the strongest voice control and longest context windows. ChatGPT is excellent for shorter tasks and has the broadest tool integration. Perplexity is the best research tool. For most marketing teams, having access to two of these covers everything.
How long should a marketing prompt be?
Working prompts for serious tasks typically run 150-500 words. The information density determines the output quality. A 50-word prompt produces a generic output. A 300-word prompt with role, task, format, voice samples, and anti-patterns produces something close to publishable. Quality scales linearly with prompt specificity.
Can AI replace a copywriter or content marketer?
It can replace junior generalists doing transactional work — basic blog posts, ad copy variants, social posts. It can't replace senior strategists who understand the business, audience, and brand voice deeply. The shift is upward: junior roles compress, senior strategic roles become more valuable.
How do I make AI sound less like AI?
Three things: (1) give it a writing sample of yours to match, (2) explicitly ban AI-tell phrases ('unlock,' 'leverage,' 'in today's fast-paced world,' 'it's important to note'), (3) edit out hedging language and replace with specific claims. AI hedges by default. Strong writing makes specific claims. The edit is what makes it usable.
John Michael Lamigo
About the Author
John Michael Lamigo
WordPress Strategist · Funnel Architect · Founder @ DigiSyn
8+ years building WordPress sites and conversion funnels for 50+ businesses across 11 industries — including work for Salt Water Digital, Growthlabz, and Dave Ramsey Solutions. Sites I've built and optimized have driven 9.28M+ Google search impressions and 56.7K+ organic clicks.