Run your Meta ads from one chat window in about an hour a week.
A written, step‑by‑step guide — since April, Meta's own official connection lets Claude (the AI chat app) into your ad account to find your best ads and draft new ones. Nothing logs in as you.
Pay by card via StripeRefunded in full, 14 days — the only condition is that you tried it. No report, no paused draft after Modules 1–2? One email to me — I'm Espen, and I answer in one business day.Everything Claude creates starts paused — it can't spend or publish. No code: if you can write an email, you can run this.$67 once, yours for good — no subscription; $97 once the first 200 are in. Runs on Claude Pro, about $20/mo; image tools optional.
It won't promise you a ROAS number. Anyone who does is selling a story.
Written by Espen Opdahl — one live Meta ad account, $102,341 of spend across 180 days, brand masked. At $2k/mo it's the same four questions on smaller totals — and one bad ad set is a bigger share of your month than of mine.
Meta lists Claude in Business Suite → Integrations. Tap it — that's Meta's page, not mine.
Leaks · Winners · New ads · Copy · Paused drafts
Everything stays paused until you approve. You switch ads on Friday. Monday, an email with the numbers.
- Find the leaks 6 months of spend, one question
- Find real winners the ads that actually sell
- Make new ads AI versions of what works
- Write the ad text 5 headlines, 5 body texts
- Draft → your call paused; you say go on Friday
official connector
still images + animated
orders, AOV
Monday email
One chat window, connected to your ad account. Five things it does every week.Solid line = Meta's official connection
What it does every week
Find the leaks
One question reads six months of your account: where the money went, and spend vs. new ads.
Find real winners
Ranked by the ads that actually sell — not the 5‑purchase ad with a 50× ROAS.
Make new ads
New AI versions of your proven ads (still images + animated), batched and named so Claude can track them.
Write the ad text
5 headlines × 5 body texts from your winning ad — every claim flagged for you to check.
Draft → approve → launch
Paused drafts in a testing campaign. You switch them on Friday. Monday, an email with the numbers.
From "which ad is actually working" to a new paused draft — in one chat.
| # | Ad | Purch. | Spend | ROAS | CPA |
|---|---|---|---|---|---|
| 1 | GOOD DAY TRUCKER – EXISTING POST – Copy | 641 | $6,330 | 5.22× | $9.88 |
| 2 | GOOD DAY TRUCKER – EXISTING POST | 537 | $5,215 | 5.40× | $9.71 |
| 3 | Carousel – Hats | 405 | $6,436 | 4.27× | $15.89 |
WINNER #2Carouselhats
Claude looks at the winning ad itself — it asks Meta for a preview and gets the picture, not the file name; most guides still say it can't. Then it writes a one‑line brief per winner (hook, visual world, offer, angle). You correct the brief; Claude makes the new versions — below: four from winner #1, same cap, new scenes. Same again for #2 and #3 → 12 new ads.
No model, no studio: winner #2 is three caps on a deck — a phone photo. Module 04 builds the same scenes from a flat product shot like that.
The top row ran; these four are new. No number ever goes under an ad that hasn't run.
Made with Claude plus an image tool — about $10–30/mo, not included in the $67. AI scenes built from the winner's own product shot. Every new ad lands paused and needs a human to approve it before it runs — Module 04 has the quality gate and how AI images get labelled; Module 01 and the Vault cover Meta's rules on AI images and where you have to say so.
| Headline | Chars | Angle |
|---|---|---|
| Buy 2, Get 1 Free | 17 | Offer |
| Sold Out Once. Back Now. | 24 | Scarcity |
| Back In Stock — Won't Last | 25 | Urgency |
| The GOOD DAY Trucker | 28 | Product |
| Shop Before It's Gone | 21 | Finality |
✓ uploaded 4 images✓ built the ad✓ created ad (paused)✓ preview
Draft created in Creative Testing → Carousel – Hats – AI v1 · status: PAUSED · preview attached. Want me to add it to the Monday email?
Winners ranked, versions drafted, nothing live until you say so.
Get instant access — $67Starts paused · 14‑day guarantee · $67 once, no subscription
$67 once. Everything opens today.
Get instant access — $67Instant access · starts paused · 14‑day guarantee · the full offer is further down
You're the founder and the media buyer and the media buyer.
usually after the shop is closed, and it's still you at 11pm.
designers are slow, real‑person videos cost a fortune, and "make more" never happens.
no weekly number of your own to say for sure why last week dropped.
Paying an agency or a buyer? Keep them — this is your own independent read on the week, not a replacement. Keep them — this is your own read on the week. Bigger, with a team? Same routine; the guide's rules for who gets access are what carry it. Same routine; the guide covers who gets access.
Built for DTC founders selling: apparel · supplements · skincare · home & kitchen · pets · food & beverage · jewelry · baby & kids · fitness gear · beauty · accessories ·outdoor · print‑on‑demand · subscription boxes · candles & fragrance · footwear · phone & tech accessories · books & courses sold via Meta · hair care · coffee · toys · art prints · stationery · sleep & wellness · and any Shopify / Woo store that runs its own ads.
Here's what's actually happening in an account like yours
Spend goes up. New ads don't keep up — every new dollar rides on the same few ads, and sales per dollar slide. It's not a targeting problem. It's more money behind the same few ads, and nobody's watching. This is the report Claude built in one question on my own account, masked:One question, my own account, masked:
[brand] — 180‑day performance
ROAS broke where new ads flattened. The account scaled from ~$500 to $12–17k a week. The Operator watches this every Monday — and fixes it every Wednesday.
Espen Opdahlwrote the guide · ran the 180 days above
refunds@claudeadsoperator.com reaches me, not a team.
See it work — 60 seconds, real account.
Connect once. Then one chat window can pull any report, find your best ads, write copy from what's already proven, and draft new ads into a testing campaign — paused, so nothing spends without you.
Five prompts, three sessions, same account as the chart above — replayed in the order they were run. Step 1 is the report at the top of this page, so the window opens on step 2.Five prompts, three sessions, same account, in the order they ran — step 1 is the report at the top of this page, so this opens on step 2. Everything Claude drafts starts paused.
Ranked by purchases, spend and ROAS together — not ROAS alone:
| Ad name | Purchases | Amount spentSpent | Purchase ROASROAS |
|---|---|---|---|
| GOOD DAY TRUCKER – existing post – copy | 641 | $6,330 | 5.22× |
| GOOD DAY TRUCKER – existing post | 537 | $5,215 | 5.40× |
| Carousel – Hats | 405 | $6,436 | 4.27× |
Runs weekly. Everything Claude creates lands paused. You approve, it launches. Measured the following Monday.
Those five prompts are the product. The guide is how you run them on your own account, in about an hour a week — 10 written modules, 45 prompts, the dashboards and the rules, already built.The guide is how you run them, every week.
Get instant access — $67Starts paused · 14‑day guarantee · same ads next week
$67 once · one payment · yours for good$67 once · one payment
The 60‑minute week.
Three short sessions. Fixed prompts. Fixed decisions. This is what replaces the daily dashboard habit.
- The Monday email, one page
- Spend / ROAS / cost per sale
- Odd swings & tired ads
- Zero‑purchase spend
- Decisions list
- No dashboard today.
- Re‑rank winners
- New versions (4–8)
- 5×5 ad text, claims flagged
- Draft paused into testing
- No dashboard today.
- Activate approved drafts
- Kill losers by rule
- Log the week
20 + 25 + 15 = 60 minutes. Three prompts you paste, three decisions you make. Module 07 puts it on your calendar. Everything Claude drafts in between starts paused.Tue, Thu and the weekend you don't open Ads Manager.
Full 180‑day re‑run, money vs. new ads, campaign‑mix review, Shopify cross‑check.
e.g. "cost per sale > target × 1.3 for 7 days and spend > $X → pause." Written down, applied by you, suggested by Claude.
Schedule the Monday email to Slack/Notion. Everything else stays one approval away.
Three sessions, sixty minutes, every week — already written down.Sixty minutes a week, written down.
Get instant access — $67Starts paused · 14‑day guarantee · first launch is Friday
No code — if you can write an email, you can run thisDoesn't work? One email in 14 days, refunded in full
The prompts don't care how many zeros are in your budget.
Same connection, same routine. What changes is which leaks it finds and how much time it hands back.
You're the whole team
- Stop guessing what to do next — the Monday email gives you a decisions list
- Make ads you can't afford a designer for, every week
- Catch the one ad set quietly burning 30% of a small budget
- One winner, 3–4 versions a week — not the twelve a $17k‑a‑month account drafts
The batch, at your size: $2k/mo is about $67 a day — you can't fund twelve tests at once, and you probably don't have three clear winners yet. Wednesday is one winner and 3–4 versions; give each about a week before you call it. The floor: 180 days at $2k is roughly $12,000 of spend, plenty for the report to read; under about $1k/mo or 90 days of history, it isn't. One ad set eating 30% of budget for two weeks is about $300 — that's what the Monday email flags. $67 once. Then Claude Pro, about $20/mo. An image tool is optional — $10–30/mo.
Where this was built
- Money vs. new ads, tracked weekly — the gap that breaks first when you scale
- Fresh ads every week without a retainer; drafts land paused in Testing
- Already have an agency or freelancer? Use the Monday email to audit them in 20 minutes
Above this, volume changes, not method: 4–8 versions is per winner, not per account — three winners is 12–24 new ads a week, all drafted paused.
Where it becomes your team's routine
- Per‑account permissions in Business Suite: read‑only for analysts, draft‑only for buyers
- One Claude Project (a saved workspace) per account — no cross‑account data; the log shows who changed what
- Daily scans for odd swings and tired ads; budget moves stay human. We teach your buyer, not replace them
The permission setup is a written page in the Vault the day you pay — which Business Suite role your analyst gets, which your buyer gets, the read‑only prompts, and the spend reconciliation check (spend = Ads Manager ±2%). Module 08 and the Vault write the team rollout, step by step and screenshotted. One $67 purchase covers you, your buyer and your analyst on your own accounts — not a licence to resell or to run an agency's other clients. Your buyer runs the Wednesday session, you read the Monday email.
| Their Business Suite access | What Claude can read | What Claude can create |
|---|---|---|
| View performance (analyst) | Spend, results, ads and images on that one account | Nothing. Read‑only prompts only — they can't draft. |
| Create ads / manage campaigns (buyer) | The same, plus previews of the ad itself | Paused drafts in the testing campaign you name |
| Full control (you) | The same | The same — the drafts still land paused |
The connector inherits the Meta role — it can never do more than that person's own login can, and the prompt library does less again: nothing in it moves a budget or a bid. Verify it before you pay: Business Suite → Integrations, where Meta lists the connector it shipped in April 2026. Where team governance lives in the guide: Module 01 (who gets which access), Module 06 (what Claude may and may not touch), Module 08 (agency partner access, one workspace per ad account when your Business Manager holds twelve, what the activity log shows when your CFO asks) — and the Vault's permission page ties the three together.
The log that answers your CFO is Meta's, not mine — Business Suite keeps it, under your own login, whether or not you ever buy this.
Not a tutorial. A system you install once — and keep.
Not paying to learn Claude. Paying to skip the work: prompts, dashboards and rules — already built on a real account, and kept current as Meta changes the connector.
10 written modules
Written step by step on a real ad account, every step screenshotted, every prompt copy‑paste. Every module ends with something working in your account, not a quiz. A step is one screenshot, one prompt, one thing done — so you can stop after five and still have finished something, then pick it up on Thursday. The screenshots carry the number of the step they belong to, which is how you find your place a week later. Here's how one is laid out.
The Operator Prompt Library
Sorted by how often you run it and what it's allowed to touch. Each prompt ships with the expected output and a "verify before acting" note, so you know what it can touch before you run it. Copy, paste, done.
45 prompts · 4 safety levelsread‑only prompts touch nothingverify: spend = Ads Manager ±2%Weekly reports
So Claude builds the same report every time: six‑month check, money vs. new ads, top winners, the Monday email.
4 weekly reports · one prompt eachTemplates
Brand Context file, workspace setup, naming convention, ad brief,claims-to-check list, weekly scorecard.
6 templates · fill in once, reuse every weekTroubleshooting Vault
"connector not enabled yet", errors, catching made‑up numbers, multi‑account, ban‑risk FAQ.
living FAQ · kept currentLifetime updates
The connector changes monthly. When it does, the prompt gets rewritten and you get an email. New reports, new modules, fixes — all yours, $0. Pay once.
The guide page and the windows above are drawn from the real pages, dashboards, prompts and files. The replay further up this page is the live version of the same account.
The full curriculum 10 written modules + a 6‑step Start Here · each one ends with a deliverable in your accountStart Here + 10 modules · all open the moment you pay
- ModuleYou leave withsteps
- StartKnow what this will and won't doStart Here · the Operator Loop, honest expectations, the stack.stack ready, workspace created6
- 01Connected to your ad account, safelyConnect & Configure · the official connection, who gets access, what to switch on first, the honest answer on bans.a connected account, Brand Context filled, naming convention set9
- 02See where six months of money wentThe 180‑Day Report · one prompt → dashboard, spend that bought nothing, money vs. new ads, checking the numbers.your report + 3 written findings12
- 03Know which three ads actually do the workFind Your Real Winners · ranked by purchases, spend and ROAS, one row per ad, reading the ad image itself, briefs.a Winner Sheet8
- 044–8 versions of each winner, weeklyThe Weekly New‑Ads Loop · versions written from the ad Claude just looked at, animate, batch to Drive, carousels, quality gate.your first batch of new ads14
- 05Copy written from what already soldCopy From Proven Winners · 5 headlines × 5 body texts, risky wording flagged.25 new lines + a wording check7
- 06Your first AI‑drafted ad, paused in a testing campaignDraft, Launch, Test · paused drafts, what Claude may and may not touch, A/B tests.one ad in testing, your risk rules written down11
- 07The 60‑minute week, on your calendarThe Weekly Operator Routine · Mon email, Wed new ads, Fri launch.scorecard + decision rules, email to Slack/Notion10
- 08Never act on a number Claude made upGuardrails & Troubleshooting · checks that catch made‑up numbers, false cause‑and‑effect, errors, multi‑account, privacy.one workspace per ad account, verification habits9
- 09Cross‑check Meta against ShopifyBeyond Meta · Shopify reconcile, Ad Library prompts, pixel health, Google pointer.a monthly cross‑check prompt7
- 10The whole loop run on a real account, written outCase Walkthrough · report → winners → versions → copy → draft → week‑1 results, every decision shown.the reference run to copy9
Start Here + ten modules. One evening to read. Then an hour a week.
Get instant access — $67Starts paused · 14‑day guarantee · same ads next week
Start Here + 10 modules · $67 once · one payment
You run them yourself. This is the routine that replaces the tutorials.You run them yourself. This replaces the tutorials.
Nobody else touches your account? This replaces the free‑tutorial pile and the evenings spent deciding what to look at — three fixed sessions a week, on your own numbers. If you do pay someone, it's your own read on their work, and it runs next to them.
| Doing it yourself now | Agency | Claude Ads Operator | |
|---|---|---|---|
| Cost | Your evenings | Whatever your retainer is, monthly, often + % of spend | $67 once (+ Claude Pro $20/mo) — runs alongside either, not instead of them |
| Independent read | None | Theirs — you get their deck | Your own, every Monday |
| New ads a week | A few a month, when you get to it | Depends on retainer | 4–8 versions per winner, every week |
| Who spots the leaks | Nobody | Their junior buyer, monthly | Monday email, every week |
| Control | Total, but manual | Low — you get a report | Total — everything paused until you approve |
| Account safety | Your own hands | Their tools, their logins — often you don't know which | Meta's official connector, under your login; you set per‑account rules |
| Time to first insight | Whenever you find an evening | 2–4 weeks onboarding | Day 1 — about 40 minutes (Module 02) |
| If it isn't working | Keep grinding | Notice period, usually 30–90 days | Do Modules 1–2, email inside 14 days — full refund |
Compared with doing it yourself, or with an agency
| You, now | Agency | Claude Operator | |
|---|---|---|---|
| Cost | Your evenings | Your retainer, monthly | $67 once + Claude Pro $20/mo |
| Independent read | None | Theirs | Your own, every Monday |
| Safety | Your hands | Their tools, their logins | Meta's connector, your login. Drafts paused. |
| First result | When you find time | 2–4 weeks of setup | Day 1 · 40 min |
Skip this one if you run the ads yourself
If you already pay an agency or a buyer: nobody fires a retainer over a $67 guide, and this doesn't ask you to. The Monday email is how you check their work in 20 minutes instead of trusting the deck: what they spent, what it returned, what went to campaigns with zero purchases tracked, and how many new ads actually shipped against the spend they added. Same three sessions, run as the audit layer — they keep buying, you stop taking the deck's word for it.
What this will not do.
Read this before you look at the price. If you were hoping for the things in the left column — a ROAS promise, an autopilot that moves budgets — this isn't your product.
What it won't do
- It won't promise you a ROAS number. Anyone who does is selling a story. What it does: hours back, more ads from what works, fewer blind spots.
- It won't move budgets or bids unsupervised — and we teach you why it shouldn't. Everything Claude creates starts paused.
- It won't put your account at new risk — Meta's official connector, your login, not some tool logging in as you. But it won't let you skip the safe‑usage rules either.
- It's not for accounts with no spend yet — the floor is about $1k/mo and 90 days of history; under that the report has nothing to read and you should keep your $67.
- It won't make the picture for you — Claude writes the brief and drafts the ad; an image tool renders it — Claude writes the brief, an image tool renders it, $10–30/mo if you want one.
- It won't decide for you. It puts the numbers and the next draft in front of you; you decide.
- It won't invent your brand voice. You fill in the Brand Context once, in Module 1, and it writes from that.
What it will do
- It will give you back the hours you spend in Ads Manager.
- It will keep new ads flowing from what's already proven — weekly.
- It will catch the leaks: zero‑purchase campaigns, tired ads, spend outrunning new ads.
- Sees the actual ad — ask for the preview and Meta's connector hands Claude the finished ad as a picture, so briefs come from the ad itself, not the file name.
- Shows its numbers, so you can check them in Ads Manager before you act.
- Works alongside a team or agency, as the audit layer.
- Runs on a normal Claude Pro plan — no code, no developer.
- Keeps every new ad paused until you switch it on.
- Stays current: when Meta changes the connector, you get the rewritten prompts.
That's the whole promise. The 14‑day guarantee is there for the case where we don't keep it.
What it can and can't touch
- Meta's own connector, under your login. Not some tool logging in as you, no password shared, nothing installed in your browser.
- 45 prompts, 4 safety levels. Read‑only prompts touch nothing; draft prompts create paused; nothing in the library moves a budget or a bid.
- One Claude workspace per ad account — no cross‑account data. Your Business Suite permissions apply: read‑only for analysts, draft‑only for buyers; the activity log shows who changed what.
- We never see your account. The connection is between your Meta and Claude accounts — we ship a written guide, prompts and templates, and never see your spend or your customers.
What you get back every week.
As they ran on the same account as the chart and the demo above. Your numbers will differ. The outputs won't.Same account as the chart and the demo above. Your numbers will differ; the outputs won't.
of zero‑purchase spend, flagged in the first email
Two campaigns had spent $25,700 in 30 days and produced zero purchases. The Monday email shows you exactly this — plus ROAS and cost per sale against your targets, tired ads flagged, and a short list of decisions.
drafted from the 3 best performers — paused, ready to test
Claude found the three ads doing the real work (38% of purchases on 23% of spend), made four AI versions of each, wrote the copy, and drafted them into the testing campaign. Nothing goes live until you say so.
approve, cut, log in one sitting — nothing spends without you
Turn on the drafts you like, pause anything that broke your written rules (e.g. "cost per sale over target for 7 days"), and log it. Monday's email tells you how the new ads did.
Friday · 15 min Your call: approve, cut, log. The drafts sit paused until you switch them on.
The Claude Ads Operator
The weekly system for running your Meta ads with Claude: 10 modules, 45 copy‑paste prompts, four weekly reports, the templates — all open the moment you pay. One payment, yours for good.
Paid once, not monthly — one day of ad budget at $2k/mo, ninety minutes of it at $30k.
Every number on this page came out of my own account, not a testimonial. The 14‑day guarantee is the rest of the proof.
Will it get my account banned? No — Meta's own connector, under your own login; nothing logs in as you.
$67 for all of it — what's the catch? No upsell: $67 is the whole product. You're buying on day one, before there is a single student result on this page — day‑one price for you, honest feedback for me.
What if I don't get to it? Say so in the same email and you're still refunded — I'm not going to audit your evenings.
Everything Claude creates starts paused · 14‑day guarantee · next week runs the same ads unless you change itStarts paused · 14‑day guarantee · same ads next week
All ten modules are written and open today. Nothing staged, no second payment, no date to wait for, and no price counting down at you. Both bonuses are included. One email inside 14 days gets your money back.
Why $67: you are buying on day one, before a single member has run this and before there is one student result on this page. That is a real risk and the price carries it — I get the first real accounts and the honest feedback, you get the whole system for one day of ad budget, and one email inside 14 days gets your money back. No timer, nothing counting down, no "only 12 left" — and no price six weeks out to hurry you, because a deadline you can't check is just a sales trick with a date on it.
Running cost$67 once. Then Claude Pro, about $20/mo. An image tool is optional — $10–30/mo. Nothing else recurring. An agency month is $2,000–$5,000; a whole year of this, Claude Pro included, is $307.
Meta's connection is newIt's Meta's, and it's young. When it changes or breaks, the fixes and rewritten prompts land in the Vault at no extra cost — that's what lifetime updates are for. Nothing it does is unpaused, so a bad week costs you drafts, not spend.Young, and Meta's. When it changes, the rewritten prompts land in the Vault free.
If it breaks on your accountThe private Q&A thread is where you ask, and I answer it myself within one business day — same inbox the refunds come to, not a support queue. If Meta pulls the connection altogether you keep everything, and the routine isn't Meta‑specific: Module 09 already points at the Google path.
14‑day "Week‑One" guarantee
Do Modules 1–2. If you don't have a full report of your account and your first AI‑drafted ad sitting paused in a testing campaign inside 14 days, email refunds@claudeadsoperator.com — full refund. It comes to me, not a team, and I answer within one business day. No forms, no calls.
And if life ate the two weeks? Didn't get to it — kids, a launch, a bad fortnight — say that in the same email and you're still refunded. I'm not going to audit your evenings. Modules 1–2 are about an hour end to end, split over two sittings — and if the fortnight got away from you, say so and you're still refunded and split into two sittings.
Is tonight $67 or $87? Modules 1–2 need Claude Pro like the rest of it — about $20/mo, no free‑tier version. So $67 plus that, unless you're on Pro already.$67 plus Claude Pro, about $20/mo — unless you're on Pro already.
And nothing is pending: all ten modules, the prompts, the templates and the Vault are written and open the moment you pay — there is no second delivery to wait for and no second payment.
Everything Claude creates starts paused · 14‑day guarantee · $67 once, no subscriptionStarts paused · 14‑day guarantee · $67 once
One payment, no subscription — and 14 days from the moment you pay to change your mindStill deciding.What you're buying.
Straight answers, in the order people ask them: is this real and is it safe, then what actually arrives for the money. Anything missing, email us; the answer goes into the FAQ.
Is this for real, and is it safe?Everything else people ask
Will connecting Claude get my ad account banned?
No — it's Meta's own connector, not some tool logging in as you. You switch it on in Business Suite, and Ads Manager sees you under your own login. Everything Claude creates starts paused; Module 1 covers the safe‑usage rules.
Check it yourself in thirty seconds: Business Suite → Integrations is where Meta lists the connector it shipped in April 2026. If it's there under your own login, it's Meta's — you don't have to take my word for it.
$67 for all of this. What's the catch?
There isn't one — here's the trade. The $67 is the whole product — no upsell; the only other cost is Claude Pro, about $20/mo. You're buying on day one, before anyone has run this and before there is a single student result on this page: you get the day‑one price, we get real accounts and honest feedback. If it doesn't work on your account, one email inside 14 days and you're refunded. No upsell: $67 is the whole product. You're buying on day one, before there is a single student result on this page — day‑one price for you, honest feedback for me.
I'm not technical. Do I need to code?
No code. You use the normal Claude app (Pro, about $20/mo), sign in with Meta once, then copy‑paste prompts from the library. If you can write an email, you can run this; if you already use Claude Code, we show that path too.
My account says the connector isn't enabled yet.
Normal — Meta is switching it on account by account. Module 1 shows where to check (Business Suite → Integrations) and what to do while you wait. You buy once and access never expires, so the whole system is sitting there ready the day Meta switches it on for you. And if your account still isn't enabled inside 14 days, that is exactly what the guarantee is for — email us and it's refunded in full.
Do you get access to my ad account or data?
No. The connection is between your Meta account and your Claude account; we ship a written guide, prompts and templates and never see your spend or your customers. We never see your spend or your customers. Module 8 covers what Meta and Anthropic each see, and how to mask brand data.
What does Anthropic keep, and does it train on my data?
The data goes to Anthropic, not to me. What it keeps, and whether it trains on it, is set on your own Claude account — Module 1 shows you where that switch is and tells you to set it before you connect anything (Module 1 shows you where). None of the 45 prompts pull customer‑level data: campaign, ad set and ad only, never your customer list or your custom audiences.
Multiple ad accounts? Google Ads?
Multi‑account setup is in Module 8 — one workspace per ad account, no cross‑client data. Core is Meta; Module 9 points to the Google path and the routine transfers.
What if Meta turns off the connector?
Then you get the updated version. Lifetime updates are part of the $67: when the connector changes, the prompts and the rules get rewritten and you get them. And the routine isn't Meta‑specific — Module 9 already points at the Google path.
Why written and not video?
Because you can search a written guide and you can't search a video. Every prompt is copy‑paste instead of pause‑and‑retype, the screenshots carry the number of the step they belong to so you find your place a week later, and when Meta changes the connector the page gets rewritten that day — a fixed video file can't be. Video walkthroughs may follow; if they do they're included, and I'm not putting a date on them.
Everything else people ask
What do I get, and is it worth it?
$67 once, no subscription, yours for good. Plus Claude Pro, about $20/mo — nothing else.
I pay the $67 — what exactly opens?
All of it, the moment you pay. You pay by card on the checkout page and your login lands in your inbox straight after — no call, no onboarding. Where it lives: a written workspace you open in a browser and log into with the email you paid with — nothing to install, readable on a phone at 11pm, and access doesn't expire. Every prompt and template in it is plain text you can copy out and keep in your own files. Your login lands in your inbox — a browser workspace, nothing to install, access never expires. Inside: the written guide in fullthe full written guide (Start Here + 10 modules, 102 steps, every stepall screenshotted),all 45 prompts, the six templates, the four weekly reports and the Troubleshooting Vault, plus the ready‑made workspace and the private Q&A thread, both included, the Vault and both bonuses. Connector not on for you yet? Module 1 covers the wait; the 14 days cover it. Nothing is staged and there is no second payment. One payment of $67, no subscription, lifetime access and updates, 14 days of guarantee running from the moment you pay. Video walkthroughs may follow later; if they do they're included, and I'm not putting a date on them.
What if it doesn't work on my account?
One email inside 14 days and you're refunded in full. The condition is that you tried it: do Modules 1–2, about an hour, split over two sittings. No report on your own account and no AI‑drafted ad sitting paused? Write to refunds@claudeadsoperator.com — it comes to me, not a team, and I answer inside one business day. No forms, no call.
And if the fortnight simply got away from you, say that in the same email and you're still refunded — I'm not going to audit your evenings.
I only spend about $2k a month — isn't this overkill?
Do the arithmetic on your own account. $67 once. Then Claude Pro, about $20/mo. An image tool is optional — $10–30/mo. At $2k spend, one ad set eating 30% of budget for two weeks is about $300 — exactly what the Monday email flags.
The batch is smaller at your size. $67 a day funds one winner and 3–4 versions a week, not twelve — give each about a week before you call it. The 12–24 figure elsewhere is per winner, from the $17k‑a‑month account.
Enough history? 180 days at $2k is roughly $12,000 of spend, plenty for the report. The floor is about $1k/mo or 90 days — under that, keep your $67.
I spend $20–50k/mo. Is 4–8 versions a week enough?
It's per winner, not per account. Three winners is 12–24 new ads a week, all drafted paused. The reason it matters at your spend: on On the account on this page, spend rose 131% in a month while new ads rose 10% — spend climbing faster than new ads is what snaps first when you scale, and it snapped at $12–17k a week. The Wednesday batch exists to keep that ratio from breaking.
We spend $100k–$200k+ with a team. Does this scale?
Yes. Per‑account rules you set in Business Suite (read‑only, draft‑only, no budget edits); Module 8 covers one workspace per ad account and activity‑log checks. One $67 purchase covers you, your buyer and your analyst, on your own accounts; it isn't a licence to resell or to run an agency's other clients. Honest limit: every figure here is from one account, $102,341 over 180 days — at your spend you'd be the first; the routine and the access rules are what carry up.
Why $67 when courses start at $447?
You're not getting a media‑buying course. Those teach buying from scratch — $447 at the low end, four figures once coaching is attached. This one assumes you already run ads and installs the routine on top. At $67 once it's under the two nearest things I found ($97/mo, or a $149 beginner workshop). It costs less because it's smaller, not because it's thin.
Can Claude see my images — and do I need an image tool?
Yes — it sees your ad pictures, not just your words. The official connector hands back the finished ad and the picture in it, across eight places it can appear. Most guides still say the opposite; we checked it against the connector's own tool list.
An image tool is optional. Module 4 demos two — Higgsfield and Google's Nano Banana, $10–30/mo. Without one Claude writes the briefs; your designer makes the visuals.
Which Claude plan — do I need Max or Team?
Claude Pro, about $20/mo, is enough. Every screenshot in the guide was taken on Pro, and the routine is one report, one Monday email and a batch of drafts a week. Move up only if you run several accounts every day; that's a convenience, not a requirement.
If your company needs a data‑processing agreement or org‑level retention controls, that's a Team/Enterprise conversation with Anthropic, not something this guide changes. The prompts and the routine are identical on any plan — I can tell you what the connector sends; I can't sign your DPA (data‑processing agreement).
How fast will I see something useful?
Day 1: your 180‑day report on screen, about 40 minutes after you connect. Day 3: the first AI‑drafted ad sitting paused in a testing campaign. Week 2: the Friday‑launch, Monday‑email routine runs itself in about an hour.
Is this a subscription?
No. One payment, lifetime access. Every update is included, and the only recurring cost is Claude Pro — about $20/mo, paid to Anthropic, not to us.
Instant access · everything Claude creates starts paused · Starts paused · one payment · 14‑day guarantee
Who's behind this.
I run one live Meta ad account — $102,341 of spend across 180 days, brand masked. Every figure on this page came out of it, and I built this system on it the month Meta shipped the connector.
What I promise is the routine: blind spots Monday, drafts Wednesday — paused — your call Friday.
What I won't promise is a ROAS number — I don't know your account, and nobody honest does. — nobody honest can.
What those six months taught me: plugging Claude into an ad account and asking it questions is not a system. Three sessions, same order, every week — that is.I built this the month Meta shipped the connector. Asking Claude questions isn't a system; three fixed sessions a week is.
No student results yet; nobody has finished week one. Until they have, the guarantee carries that risk, not you.
— the founder · this address reaches me, not a teamemail me directly
- BuiltApr 2026
- Evidenceone ad account · 180 days
- Hiddenbrand + account name, nothing else
- Claimeda routine — never a ROAS
- Not yetstudent results
- Refund14 days, week one
There are no testimonials on this page, and none will go up until they come with the quote verbatim, the member's own week‑one report, and their name and store with permission. That is the whole standard, and every figure you have read on this page was held to it.No testimonials here. None go up without the quote verbatim, their own week‑one report, and their name — with permission.
No quote here will ever carry a ROAS claim.
- Day 1Your 180‑day report on screen, about 40 minutes after you connectYour 180‑day report on screen, ~40 min after you connectsee it
- Day 3Your first ad drafted from your own winners — pausedsee it
- MondayOne email of what changed, every week after that, every weeksee itsee it
- Day 14$67 refunded if week one didn't deliverthe terms
Your ad account. One chat window. One hour a week.
Connect today. Have your 180‑day report and your first AI‑drafted ad before the weekend.
10 modules45 prompts4 weekly reports6 templatesTroubleshooting VaultLifetime updates2 founding bonuses, free
Everything Claude creates starts paused · 14‑day guarantee · next week runs the same ads unless you change itStarts paused · 14‑day guarantee · $67 once
Start Here + 10 modules · $67 once · one payment, no subscription, yours for good
$67, once. One leaked ad set costs you $300 at $2k/mo before you catch it; a Meta‑ads course is $447–$2,500; an agency month is $2,000–$5,000 — and the same ads run again on Monday either way.
The next minute you pay $67 once and the whole written guide — all ten modules, the prompts, the templates — opens straight away; nothing staged, no second payment. The first 40 minutes — connect the account, read your own 180‑day report. The 14 days after — do Modules 1–2; no report, no paused draft, one email and it's refunded.