Better prompts for AI
This prompt guide is based on Stop settling for "this isn't what I wanted" from AI, featuring Pam Didner, published on the AI Lab by ActiveCampaign.
Get the prompt template
How to use these prompts
These are ready-to-use prompts pulled from the workflow Pam Didner teaches in her AI trainings for Southwest Airlines, Intel, and the U.S. Army. Copy, paste, and swap in the details where you see [BRACKETS]. Every AI tool and model behaves a little differently, so treat what comes back as a starting point — review the output, then re-prompt to sharpen it.
Prompt 1: Diagnose a declining email campaign
Best for: troubleshooting a specific campaign metric that has drifted over time, and getting testable experiments you can run next quarter.
Use with: any LLM (ChatGPT, Claude, Gemini)
Prompt 1
My monthly B2B email campaigns targeting [JOB_TITLE] in [REGION] have experienced a steady decline in open rates — from [XX]% last year to [XX]% today. These emails are newsletters that include industry insights and case studies, sent via [PLATFORM] on the first Tuesday of each month. Subject line examples include: [2–3 EXAMPLES]. I’ve attached a file containing monthly subject lines and sub-subject lines, along with performance indicators. Please analyze the file, identify possible reasons for the decline, and recommend specific, testable strategies I can implement over the next quarter to improve open rates.
Variables to fill in:
- [JOB_TITLE] — your target persona (e.g., “VP of Operations”)
- [REGION] — the region you sell into (e.g., “North America”)
- [XX]% — the old and current open rates
- [PLATFORM] — your sending tool (e.g., ActiveCampaign)
- [2–3 EXAMPLES] — recent subject lines so the model sees your style
What to expect: a list of possible diagnoses grouped by theme (subject line wording, send-time cadence, segment drift) and 4–6 experiments you can run this quarter. Attach your real performance file — without it, the analysis will stay generic.
Follow-up prompt
I’ve already tried most of the strategies you suggested. Can you now provide less conventional or “novice” ideas — things that may not be widely used in B2B email marketing but could differentiate my campaigns and capture attention?
Pam often finds the “novice” follow-up is where the interesting ideas live. The first pass gives you table-stakes advice; the second pass pushes past the obvious.
Prompt 2: Understand when buyers use AI in their journey
Best for: figuring out where in the decision-making process your ICP is consulting AI, so you know which pages and messages have to be AI-ready.
Use with: any LLM
Prompt 2
Here’s my ideal customer profile [PASTE_ICP]. What role does AI likely play in their early evaluation process? At what stage of decision-making would my brand need to show up?
Variables to fill in:
- [PASTE_ICP] — your actual ICP document: company size, industry, role, jobs-to-be-done, buying triggers
What to expect: a mapped journey with the specific stages (problem awareness, vendor shortlisting, deep evaluation) where AI likely enters the picture, plus recommendations for where your brand needs to be visible. Use the output to audit which pages your sales team should prioritize, and share the findings with sales so outreach timing matches buyer behavior.
Prompt 3: Understand how buyers use AI alongside search
Best for: reverse-engineering the actual questions buyers ask in Google and ChatGPT about your category, so you can write content that answers both.
Use with: ChatGPT (needs to be able to reason across your site and competitors)
Prompt 3
Here’s information from my website about my company [PROVIDE_INFO], as well as information from my competitors [PASTE_COMPETITOR_CONTENT]. Show me how [INSERT_POSITION_OR_ICP] might use Google and ChatGPT together to research [YOUR_PRODUCT_CATEGORY]. What questions would they ask in each tool?
Variables to fill in:
- [PROVIDE_INFO] — homepage copy, product pages, positioning statements
- [PASTE_COMPETITOR_CONTENT] — 2–3 competitors’ equivalent pages
- [INSERT_POSITION_OR_ICP] — the specific role you’re modeling (e.g., “head of IT at a mid-market bank”)
- [YOUR_PRODUCT_CATEGORY] — your category (e.g., “marketing automation platforms”)
What to expect: two parallel lists of questions — one for Google (shorter, keyword-style) and one for ChatGPT (longer, conversational). One of Pam’s clients, a technical products company, used this to learn which technical challenges their ICPs were researching in AI and then repositioned the product page accordingly.
Prompt 4: Reverse-engineer the prompts driving ChatGPT referral traffic
Best for: finding out which prompts buyers are typing into ChatGPT to land on your high-traffic pages, so you can expand that content intentionally.
Use with: any LLM
Prompt 4
Here are my top 5 traffic pages from ChatGPT referrals [LIST_URLS]. Suggest the 5–8 likely prompts buyers are using to arrive at these pages.
Variables to fill in:
- [LIST_URLS] — the URLs that show ChatGPT as a referrer in your analytics
What to expect: a list of plausible buyer prompts for each page. Use it to spot where you’re accidentally ranking in AI results and double down. Pam’s evergreen post on top B2B marketing conferences came directly from this prompt — she saw the traffic, realized buyers were asking about conferences, and turned it into a page she updates every year.
Prompt 5: Audit your homepage for Answer Engine Optimization
Best for: getting a prioritized AEO punch list for your homepage without paying for a full SEO audit or rebuilding the site.
Use with: ChatGPT (can fetch or review the page directly)
Prompt 5
Evaluate the homepage of my website: [INSERT_URL]. Analyze it specifically for Answer Engine Optimization (AEO) and AI Search Optimization. Look at elements such as headline clarity, metadata, structured data, keyword usage, content depth, FAQ coverage, and overall search intent alignment. Identify gaps where my homepage may not perform well for AI-driven search engines like ChatGPT, Gemini, or Perplexity.
Then, provide a prioritized list of specific recommendations I should implement (e.g., schema markup, conversational content, query- style headings, better internal linking, etc.) to increase my chances of being surfaced as an authoritative answer in AI search results.
Variables to fill in:
- [INSERT_URL] — the homepage (or any page) you want evaluated
What to expect: a gap analysis with prioritized fixes. Pam uses this to catch where a page fails to explain something clearly or lacks credibility markers like social proof. Treat the output as a backlog — tackle the top 3 recommendations first, then re-run the prompt after edits to see what changed.
Prompt 6: Draft a buyer persona from your own material
Best for: producing a first-draft persona when you don’t have the time or budget for full interviews, so the team has a shared mental picture to pressure-test.
Use with: any LLM (pair with a custom GPT or project for reuse)
Prompt 6
FIRST, PROVIDE CONTEXTUAL INFO LIKE:
- Company website pages
- Company blogs
- Product features and pages
- Customer journey notes
- Branding and messaging guides
- Customer testimonials and case studies]
Prompt: Create a customer persona profile for a [JOB_TITLE] at [COMPANY] as part of our marketing plan.
Details to include:
- Assign the persona a realistic name.
- Approximate age: around [NUMBER].
- A quasi job title (not the formal corporate title, but a descriptive role).
- Family background: [PROVIDE_DETAILS]
- Location: [PROVIDE_LOCATION]
Persona structure:
- Name, Quasi Job Title
- 4 personality traits
- One summary statement capturing their needs for [PRODUCTS_OR_BRANDS].
- 2–3 short paragraphs describing how they use [PRODUCTS_OR_BRANDS].
- Goals/Aspirations: 4–5 bullet points
- Challenges/Pain Points: 4–5 bullet points
Variables to fill in:
- [JOB_TITLE], [COMPANY], [NUMBER], [PROVIDE_DETAILS], [PROVIDE_LOCATION], [PRODUCTS_OR_BRANDS]
What to expect: a named, structured first-draft persona with goals and pain points. Pam treats the AI output as v1 — she adds Google research, hands it to the client for edits, and then re-uploads the finalized version back to the AI tool so future prompts can reference it. Once it’s final, loop the persona back into your tool as context — future prompts that reference it will inherit the detail without you pasting it again.
Tips for better results
- Prompting is re-prompting. The first draft will almost never be perfect. Treat each output as a starting point and say the same thing differently until you get closer.
- Add a self-check step. Ask the model to explain how its output aligns with your brand guidelines, or rate itself on tone, vocabulary, and CTA style. Catching “misses” before you review saves a pass.
- Adapt prompts per platform. A prompt that works for a LinkedIn post will not land the same way for TikTok copy. Rewrite for the format instead of copy-pasting across channels.
- Always fact-check. AI hallucinates and invents stats. Review every output, especially anything you’re about to publish.
Related
More data from the AI Lab.