My AI prompt secret library
People often ask me how I manage to run SEA campaigns, SEO strategy, paid social, content and website projects for multiple clients without losing my mind. The honest answer is that I do not do it alone any more. AI tools have become a genuine team asset, but only because I stopped using them like a search bar and started using them like a sharp, slightly overconfident junior analyst who needs clear instructions and a healthy dose of fact-checking.
This is the library of prompts I actually use, week in and week out, across SEA, SEO, paid social, social media and website strategy. Not generic templates pulled from a listicle. The specific structures that save me hours, surface genuinely useful insights and, just as importantly, help me figure out how much to trust what comes back.
Why does prompt structure matter more than the question itself?
A vague prompt produces a vague answer. That is the single biggest lesson after two years of using AI daily in client work.
The most reliable structure I use is what some call the Context, Task, Format formula. You tell the AI who you are and who your audience is, what exactly you want it to do, and how you want the output structured. The more specific you are, the more useful and on-brand the output becomes. “Help me with SEO” produces generic advice. “Act as an SEO strategist working with a B2B SaaS client targeting enterprise buyers in the UK; build a content brief for a comparison page targeting [keyword], including search intent, internal linking strategy and AEO and GEO optimisation” produces something I can actually use.
Get a free audit and find out exactly what to fix first.
What are my go-to prompts for SEA and Google Ads work?
These save me the most time on a weekly basis.
For search term analysis: “Act as a Google Ads strategist. Review this list of search terms [paste search term report] and categorise them into high-intent, low-intent and irrelevant. Recommend which terms should be added as negative keywords and explain your reasoning for each.”
For budget allocation modelling: A multi-step approach works better than asking for everything at once. I break complex analysis into prompt chains. First, I ask for current performance: “Calculate ROAS by channel including Google Ads, Meta Ads and LinkedIn Ads, using last-click attribution with a 30-day window, based on this data [paste data].” Then I follow up with a second prompt building on the first response, asking for incremental spend scenarios. Decomposing a complex question into sequential, focused prompts consistently produces better analysis than asking for the full answer in one go.
For ad copy testing: “Write five Google Search ad variations for [product/service], targeting [audience], emphasising [key benefit]. Include headlines under 30 characters and descriptions under 90 characters. Highlight which variation tests the strongest emotional hook versus the strongest rational argument.”
How do I use AI for SEO and content strategy?
This is where the time savings are most dramatic, because SEO is largely repeatable process work.
SEO is roughly 80% repeatable process. Keyword research follows patterns. Content briefs have templates. Technical audits run the same checklist. AI does not replace SEO thinking, but it eliminates the grunt work that eats an entire afternoon.
My most-used SEO prompts:
- Content brief generation: “Build a content brief for a blog article targeting [keyword], for an audience of [description]. Include search intent classification, secondary keywords, recommended word count, heading structure and an AEO-optimised answer for the opening paragraph.”
- Three-month content planning: “Build a 3-month content plan for a [industry] business targeting [location]. Include primary keywords, secondary keywords, search intent, internal linking strategy, and optimisation for AEO and GEO.”
- Technical audit summaries: “Summarise this Search Console export [paste data] and identify the three highest-priority technical issues based on traffic impact and ease of fix.”
- Meta tag testing: “Write five title tag and meta description combinations for this page [paste URL or content], optimised for both click-through rate and AI citation potential.”
What about paid social and content creation?
Brand voice consistency is the single hardest thing to get right with AI-generated social content, so this is where I am most deliberate with my prompts.
Making AI-generated social content sound authentic requires specific techniques. I always provide examples of previous high-performing posts that captured the right voice, then instruct the AI to write conversationally, with varied sentence lengths and genuine personality rather than corporate polish.
My standard social content prompt: “Here are three examples of our best-performing LinkedIn posts [paste examples]. Write a new post about [topic] in the same voice, using contractions, varied sentence length and a conversational tone. Avoid generic marketing phrases and clichĂ©d openings.”
How do I use AI for website and landing page work?
This is where conversion-focused thinking matters more than pure content generation.
My go-to prompt: “Act as a conversion-focused copywriter. Write a landing page for [service] targeting [audience]. Include headline variations, value propositions, objection handling, trust signals and a strong, specific call to action.”
I also use AI to stress-test existing pages: “Review this landing page copy [paste content] and identify where the message does not match what a paid search ad promising [specific offer] would lead a visitor to expect.”
How do I challenge AI engines instead of just accepting the first answer?
This is the habit that has saved me from the most embarrassing mistakes, and it is the one most people skip.
I never accept a first response at face value, particularly for anything involving statistics, market data or strategic recommendations. My standard follow-up prompts:
“What assumptions are you making in this answer? List them explicitly.”
“What is the strongest counter-argument to the recommendation you just gave me?”
“Where might this answer be outdated or based on incomplete information? Be specific.”
“On a scale of confidence, how reliable is this specific claim, and why?”
These follow-ups consistently surface caveats, outdated assumptions or oversimplifications that the first answer glossed over. AI tools are confident by default. Challenging that confidence is part of the job now, not an optional extra step.
How much should you actually trust AI-generated answers?
This is the question every marketer using AI daily needs a clear answer to, and mine has become more disciplined over time.
I treat AI output on a sliding scale. For repeatable structural tasks, content briefs, ad copy variations, meta tag drafts, I trust the output heavily and edit lightly. For anything involving current statistics, market benchmarks or platform-specific policy details, I verify independently before using it in client work, because models can confidently state outdated or incorrect figures.
Ranking success still requires human expertise in strategy, understanding ranking factors beyond content, and ensuring genuine value and originality. AI helps execute best practices more efficiently. It does not replace strategic judgement.
How do you build your own prompt library?
Start simple. Every time you write a prompt that produces a genuinely useful result, save it, with the context that made it work. Over time, this becomes a reusable system rather than a one-off effort.
According to HubSpot’s 2026 AI in Marketing research, teams using structured prompt libraries reduce content production time significantly while maintaining stronger brand consistency than teams improvising prompts case by case. And according to Search Engine Land’s AI prompting guidance, the strongest results consistently come from specificity: defining audience, format, tone and intent clearly, rather than relying on AI to infer what you actually need.
AI has not replaced my expertise across SEA, SEO, paid social and website strategy. It has removed the repetitive grunt work that used to eat entire afternoons, freeing up time for the strategic thinking that actually moves campaigns forward.
The prompts matter. But the habit of challenging the output matters more. Treat every AI response as a strong first draft from a capable but occasionally overconfident analyst, not as a finished answer.
Caroline is a Performance Marketing Consultant, specialising in paid media strategy, SEA, paid social and lead generation for B2B and B2C brands.
View All Articles