Facebook Ads, Run by Talking
Run real Facebook ad campaigns using nothing but plain-English conversations with Claude. No dashboards. No jargon. Just the prompts you copy, paste, and send — plus the handful of rules that separate people who make money on Facebook from people who just spend it.
The story, in plain English
A business wanted signups for a free online workshop. They were running a pile of Facebook ads — some spending a lot and getting nothing, others not spending at all, and the dashboard was a wall of numbers. So instead of squinting at it, the owner talked to Claude: “What’s running? What’s wasting money? What’s actually getting me signups?” Claude read the account, did the math, and answered in sentences. Then, step by step: turn off the waste, pick one number that means success, run a fair race between two ways of targeting, and pour money into the winner. Signups came in at about a fifth of what the losing campaign had been paying — and the whole thing ran through a conversation. Here’s how you do it yourself.
The mental model
You are the boss. Claude is your media buyer — the expert who actually touches Facebook. You decide the offer, the budget, and every big yes/no call (especially spending more money). Claude does the setup, the reading, the math, the changes, and the watching — and tells you the trade-offs before doing anything you can’t undo.
One habit to insist on, always: “Show me the real number you just pulled.” Early on it’s easy for anyone — human or AI — to say “done!” when a change didn’t fully go through. The prompts below build that check in.
Get set up — connect Facebook to Claude
One-time, about ten minutes. A “connector” is just a permission slip that lets Claude talk to Facebook Ads on your behalf.
- What you need: a Claude account with Cowork (a paid plan), a Meta Business account with an ad account where you are an admin, a payment method on the account — and ideally a Facebook Page and pixel already created.
- Open connectors: in Claude, click “+” at the bottom of the message box → Connectors → Manage connectors.
- Add Meta / Facebook Ads: search the directory, click Connect.
- Log in on Facebook’s own page and grant ad-management access for the Business, ad account(s), and Page(s) you want managed. Claude never sees your password; disconnect any time.
- Toggle it on for your chat — connecting and using are two different switches. If Claude “can’t see your ads,” check this first.
- Prove it works — send the connection test prompt below before trusting it with money.
Gotchas: your ad account reports in ONE currency and timezone (“today” means the account’s day, not yours) · the pixel is your scoreboard — if it’s missing or misfiring, nothing works · read-but-can’t-change means you’re not an admin · a nervous beginner’s best setting is “Needs approval” on anything that spends money.
The 10 golden rules
Every one was paid for with real money.
- Start BROAD and let Facebook do the hunting. Don’t hand-pick who sees your ad. Give Facebook a wide audience and good creative, and its algorithm finds the buyers. Broad beat hand-picked targeting 5× in our own test.
- Test several creatives, not one. You are a bad judge of your own ads — the one you love often flops, the throwaway wins. Launch 3–5 images/videos and let the audience vote with clicks.
- Only chase a goal Facebook can actually see. Facebook optimizes toward events it can detect (a signup, a purchase). If the goal isn’t actually tracked, it gets confused and barely spends. Confirm tracking before you spend.
- Pick ONE number that means “success.” For us it was cost-per-signup. One number stops every wobble from turning into a panic.
- Give it 2–3 days before you judge it. Don’t fiddle. New campaigns go through a learning period where costs bounce and mean nothing. Every budget change resets that clock. Set it, walk away, come back in two days.
- Scale up slowly — nudge, don’t yank. When something works, raise the budget 20–30% at a time, not double. Big jumps shock a winner back into learning.
- Kill your losers without guilt. After a fair chance, turn clear losers off. No sentiment — the money is better spent on the winner.
- Match your ask to how warm the audience is. Cold audiences do a small, easy thing (register for something free). They will not fill a long form. Save big asks for people who know you.
- Don’t confuse “busy” with “working.” A page can be buzzing from email and word-of-mouth. Always ask for the results the ADS drove — not the total.
- At the end, re-target the warm crowd. Before a deadline, run a cheap “last chance” ad at visitors who didn’t convert (excluding those who did). It converts far cheaper than cold.
The prompts — copy, fill the [BRACKETS], send
Work the first six top to bottom; keep the rest handy for day-to-day.
Confirm you’re connected to my Facebook ad account. Tell me the ad account name, its ID, what currency and timezone it’s set to, and how many campaigns exist. If you can see that, we’re good to go.
I want to run Facebook ads and I’m not technical — you’ll handle the Facebook side and explain everything to me simply. First, look at my Facebook ad account and tell me, in plain English: which campaigns are currently running, how much each is spending per day, and whether any of them look like they’re wasting money. Don’t change anything yet — just show me what’s there.
Based on what you found, tell me which campaigns are old, inactive, or clearly wasting money and should be turned off. List them with a one-line reason each, then pause them for me once I say go. I want a clean slate with only the ads that matter to my current goal.
My goal is to get people to [SIGN UP FOR MY OFFER] on this page: [YOUR PAGE URL]. Before we spend a dollar, check my Facebook pixel and tell me in plain English whether a [signup] is actually being tracked on that exact page. We should only optimize for something you can confirm is really happening. If it’s not being tracked properly, tell me what to fix first.
Here’s my plan: I want [signups], my total budget is [$X] per day, and success to me means getting a signup for [$Y] or less. Set that one number — cost per signup — as the way we judge everything. From now on, whenever I ask how things are going, tell me that number first.
Set up a simple test for me. Use my [3–5 creatives] I’m providing. I want two versions running side by side with the SAME goal (a signup), so the only difference is who we target: 1) A BROAD audience — let Facebook find the buyers itself. 2) A NARROW audience — people with specific interests related to [MY CUSTOMER]. Point every ad at [MY PAGE URL], split my budget across them, and add proper tracking tags to every ad so we can tell later exactly which ad brought each signup. Walk me through what you set up before it goes live.
How are my ads doing today so far? For each campaign, give me: how much it’s spent, how many signups it got, and the cost per signup. Put the best performer first and tell me plainly if anything looks off.
Is anything wasting my money right now — spending a lot with little or nothing to show for it? If so, tell me which, how much it’s burned, and whether you’d pause it. Don’t change anything until I say go.
[Campaign name] seems to be working well. Should I put more budget into it? If yes, how much more can I add without shocking it back into the learning period? Give me the safe amount.
The numbers today look [worse/weird] to me and I’m tempted to change something. Before I do — is this just normal early bouncing / a weekend dip, or is it a real problem? Tell me honestly whether to act or wait.
Give me a quick end-of-day summary: total spent, total signups, cost per signup, which campaign won today, and one recommendation for tomorrow. Keep it short.
It’s been a few days and the campaigns are out of the learning period. Compare my broad vs narrow test on cost per signup. Tell me clearly which one won, recommend moving budget into the winner and pausing the loser, and explain the trade-off in one line before we do it.
We’re [3] days from my deadline. Set up a “last chance” ad aimed only at people who already visited my page or engaged with my ads — but exclude anyone who already signed up. Schedule it to run only from now until [deadline], and keep it in its own campaign so it doesn’t mess with the others.
[Campaign name] isn’t spending much. Walk me through the likely reasons in plain English — is it still learning, is the goal something Facebook can’t track, is the audience too small, or did a budget change accidentally pause it? Check each and tell me what you find and how to fix it.
The whole game on a napkin
Look (what’s there, change nothing) → Clean (pause the dead weight) → Check tracking (is the signup really tracked?) → Race (broad vs narrow, same goal, 3–5 creatives) → Wait 2–3 days, then back the winner. You don’t need to master Facebook. You need five prompts, ten rules, and the patience to wait two days.
Want this set up for your business — by name?
Take the 5-minute diagnosis and meet your first AI employee. If paid ads are your highest-leverage fix, your report onboards an AI Media Buyer around your offer, your page, and your number — with every prompt above personalized and every dollar held for your yes.