Most people use AI on Meta ads for exactly one thing: writing captions. That barely scratches the surface. Used well, a model like ChatGPT or Claude becomes a media-buying co-pilot — it plans campaign structure, builds audience angles, drafts ten copy variations before your coffee is cold, reads a messy Ads Manager export and tells you what to pause, and turns a week of numbers into a client report in seconds.
Below are 30 prompts we actually use, grouped by the job you're trying to do. Copy them, swap in your own product and numbers, and they work. One rule before you start, and it matters more than any single prompt:
How to use this guide
The prompts get more powerful as you feed them context. A prompt like "write a Facebook ad" gives you generic mush; the same request with your product, audience, offer, and objective gives you something you can almost ship. Wherever you see [square brackets], replace them with your real details. If you don't have an ad account yet, the planning, audience, and copy prompts still work as training reps — treat them as the way a media buyer actually thinks, out loud.
Campaign planning & strategy
Start here whenever you're setting up something new. These prompts turn a vague goal ("get more leads") into a structured plan you can build in Ads Manager.
“Act as a senior Meta media buyer. I sell [product/service] to [audience] in [location], with a budget of ₹[amount] per day and a target cost per lead of ₹[amount]. Recommend a campaign structure — objective, number of ad sets, and audience split — and explain the reasoning behind each choice in plain language.”
“I'm choosing between a Sales (conversion) campaign and an Advantage+ Sales campaign for [product]. Explain the difference for a beginner, list when each one is the right call, and tell me which you'd pick for my situation: [describe budget, audience size, and how much conversion data my pixel already has].”
“Give me a 3-stage funnel plan for [product] on Meta — cold (awareness), warm (engagement/traffic), and hot (retargeting) — with a suggested budget split, the objective for each stage, and one creative angle per stage.”
“I have ₹[total budget] to spend over [number] weeks. Build me a week-by-week testing plan: what to test first (creative, audience, or offer), how long to let each test run before judging it, and what result would tell me to scale vs. kill.”
Audience & targeting angles
Meta's targeting has narrowed over the years, so the winning ideas now come from angles — the different reasons different people buy — more than from obscure interest checkboxes. AI is genuinely good at brainstorming these.
“List 10 distinct customer personas who might buy [product], and for each one give me: their core pain, the emotional trigger that makes them act, and one ad angle that would speak directly to them.”
“For [product], suggest 8 interest and behaviour targeting ideas I could test on Meta, and — just as important — tell me which ones are probably too broad or too obvious to bother with, and why.”
“Give me a retargeting audience plan for [product]: which website and engagement events are worth building custom audiences from, how to sequence the messaging for people at each stage, and which audiences are too small to be worth a separate ad set.”
Ad copy & creative
This is where AI saves the most time — but only if you brief it like you'd brief a junior copywriter. Give it the angle, the constraint, and one example of the voice you want.
“Write 5 primary-text variations for a Meta ad selling [product] to [audience]. The angle is [pain point / benefit]. Keep each under 125 characters, lead with a hook in the first line, and end with a clear call to action. Indian English, warm and direct — not corporate.”
“Write 10 headline options (under 40 characters each) for this ad. Mix curiosity, benefit, and urgency angles, and label which is which so I can test one of each type.”
“Turn this one testimonial into 3 ad angles: [paste testimonial]. For each, give me the hook line and the emotional promise it leans on.”
“Give me 5 video hook scripts (first 3 seconds) for a Reel promoting [product]. The job of each is to stop the scroll — pattern interrupt, bold claim, relatable problem, question, or visual surprise. Keep them speakable in one breath.”
“Rewrite this ad copy for 3 different awareness levels: someone who's never heard of the problem, someone actively looking for a solution, and someone comparing us to a competitor. [paste copy]”
Reading your numbers (analysis)
Once a campaign is live, the skill that separates real media buyers from button- pushers is diagnosis. Export your data from Ads Manager, paste it in, and let AI help you see the story — then verify it yourself.
“Here is my Meta Ads data for the last 7 days [paste table: campaign, spend, impressions, CTR, CPC, purchases, ROAS]. Summarise how the account is doing, flag anything that changed significantly vs. what I'd expect, and list the 3 things I should look at first.”
“For this campaign, my CPM is ₹[x], link CTR is [y]%, and cost per purchase is ₹[z]. Walk me through — like you're teaching a junior — whether the problem is the audience, the creative, or the landing page, and how I'd tell the difference.”
“Explain these metrics to me as if I'm new: hook rate, hold rate, CTR, CPM, CPC, CPA, and ROAS. For each, tell me what a "healthy" number roughly looks like and what a bad number is usually telling me.”
“Compare these two ad sets and tell me which is genuinely winning and which just got lucky on low spend: [paste both]. Explain how much spend I'd want behind each before I trust the result.”
Creative fatigue & scaling decisions
The two questions every buyer asks daily: "is this ad dying?" and "can I safely spend more?" AI is a good second opinion on both — with your numbers in front of it.
“Here's an ad that's been running for [x] days: frequency is [y], CTR has moved from [start]% to [now]%, and CPA has gone from ₹[start] to ₹[now]. Is this creative fatigue, and should I refresh the creative, cut spend, or keep going? Explain your reasoning.”
“Give me a simple checklist to diagnose creative fatigue on Meta — the exact signals to watch (frequency, CTR decline, rising CPM/CPA) and the rough thresholds where each becomes a real warning, not just noise.”
“This ad set has hit a stable, profitable cost per result for [x] days at ₹[budget] a day. Give me a scaling plan: how much to raise the budget at a time, how long to wait between increases, and the signs that would tell me I've scaled too fast.”
“My best creative is fatiguing. Give me 5 fresh variations of the same winning angle — different hook, format, or framing — so I can refresh without throwing away what's already working.”
Client reporting & freelancing
If you run ads for clients — or want to — reporting is where hours quietly disappear. These prompts turn raw numbers into something a non-marketer client actually understands.
“Turn this month's Meta Ads data into a short client report a non-technical business owner can understand: what we spent, what we got (leads/sales), how it compares to last month, and what we're changing next month. No jargon. [paste data]”
“A client is worried their cost per lead went up this month. Using this data, write me a calm, honest explanation of what happened and what we're doing about it — without over-promising. [paste data]”
“Draft a proposal outline for a ₹[amount]/month Meta Ads retainer for a [type of business]. Include what's in scope, what results are realistic to promise (and what isn't), and how I'd justify the fee.”
Learning & interview prep (for students)
This is the section the agency guides skip — because it's for people building the skill, not just using it. If you're studying Meta ads, AI is the most patient tutor you'll ever have.
“Quiz me on Meta Ads like an interviewer would for a junior performance-marketing role. Ask one question at a time, wait for my answer, then tell me if I'm right and what a stronger answer would include.”
“I ran a practice campaign and got these results: [paste]. Pretend you're a hiring manager reviewing my portfolio. What would impress you, what would worry you, and what questions would you ask me about these numbers?”
“Explain the difference between CBO (Advantage campaign budget) and ABO (ad-set-level budget) with a real example, then give me a scenario and ask me which I'd choose and why.”
“Give me a 2-week study plan to go from "I understand the Ads Manager interface" to "I can plan, launch, and read a basic campaign" — with one small hands-on task per day.”
Prompts that touch your live account
Some AI tools connect directly to your ad account and can read live numbers or draft changes for you to approve. These are powerful and worth learning — but they're also where discipline matters most. Every one of these should end with you approving the change in Ads Manager.
“Pull my account performance for the last 7 days vs. the previous 7, and tell me which campaigns changed significantly in spend, ROAS, or cost per result — then stop and let me decide what to do. Don't change anything.”
“Find every ad set that has spent more than ₹[amount] in the last 7 days with zero purchases, list them for me with their numbers, and propose which to pause — but wait for my confirmation before pausing anything.”
“Draft (don't publish) 3 new ad variations for my best-performing ad set based on what's already working, so I can review and launch them myself from Ads Manager.”
How to get much better answers from AI
The gap between a useless AI answer and a genuinely sharp one is almost never the model — it's the prompt. Four habits do most of the work:
- Give it a role and a goal."Act as a senior Meta media buyer" plus a clear objective beats a cold question every time.
- Feed it real numbers and context. Your product, audience, budget, objective, and actual metrics. Vague in, vague out.
- Constrain the output."5 variations, under 125 characters, Indian English, one hook per line" gives you something usable; "write an ad" doesn't.
- Ask it to show its reasoning."Explain why" turns a black-box answer into something you can learn from — and catch when it's wrong.
The honest limits
AI will confidently state numbers that don't match your Ads Manager, invent benchmarks, and miss context only you have — a seasonal spike, a landing-page bug, a client's real margin. It doesn't know your account the way you do. Use it to think faster and to catch things you missed, never as a reason to stop checking. And never paste a client's sensitive data into a tool you haven't confirmed is private.
Where the judgment comes from
Every prompt here assumes one thing: that you can read the answer critically. That's not something you get from a prompt list — it comes from running real campaigns, making real mistakes with real (if small) budgets, and having someone experienced tell you what you missed. That's exactly how the performance-marketing track works at Unlearn Academy — students run live Meta campaigns on real client briefs, with mentors who buy media every day, so the judgment is built before the job interview. You can see how that's structured on the Digital Marketing course page.
Frequently asked questions
Sources
About the author
Hassan Rawther
Founder, Adsomia & Unlearn Academy
Hassan has spent 12+ years in digital branding and paid advertising, and founded Adsomia — a Kerala-based agency that is a Meta Business Partner and Google Partner — before starting Unlearn Academy to train the next generation of marketers the same way his own team works. He leads the performance-marketing track at Unlearn, where students run real Meta campaigns on live client briefs. Every prompt in this guide comes from workflows his team uses in production, not from theory.
More about the team