Give Claude Control of Your Social Media

Why your content calendar should be Claude's problem

Most content advice tells you to "post consistently" and "watch your analytics". So you post, you glance at the numbers, one reel randomly does 10x the others, and you have no idea why — so you can't do it again. That's not a strategy, that's a poker machine.

Here's the shift: stop asking Claude to write content and start asking it to decide what content is worth making. Claude is genuinely good at pattern-finding — it just needs real data to find patterns in. Give it two feeds:

😵‍💫 You right now

Stares at a blank notes app on filming day
Posts, hopes, checks views obsessively
Can't say WHY the good one worked
Sees a competitor blow up, feels behind

🧠 You with this system

Opens a ready-made film list on Monday
Every idea is backed by real numbers
Knows the format, hook and structure before pressing record
Competitor breakouts become YOUR briefs

The receipts

This is the system I actually run — not a thought experiment. My accounts passed 60,000 followers, and most of that growth happened in the last six months, after I handed the "what should I film?" decision to Claude. I still film everything myself. Claude just makes sure I never film a guess.

The system has two data feeds and one weekly habit. Both feeds take about ten minutes each to set up, and neither needs a single line of code.

Feed it your own wins — connect Windsor

Windsor AI is a connector tool that pulls your real performance data — Instagram, TikTok, Facebook, YouTube — into one place Claude can read. Views, reach, shares, saves, watch time, per post. It has a 30-day free trial, no card drama, and setup is "log in, connect your accounts, done".

1

Create a Windsor account

windsor.ai → free trial → connect your Instagram (and TikTok if you post there). It starts pulling your post history straight away.

2

Export your last 90 days

Pull a CSV of your posts with views, reach, saves and shares. (If you use Claude with connectors, you can wire Windsor in directly and skip the manual export.)

3

Hand it to Claude with this prompt

Attach the export (or point Claude at the connection) and paste the prompt below. This is the reverse-engineering step — it's where generic advice dies.

Here's my last 90 days of social media performance data. Reverse-engineer what's actually working on MY account — not general best practice. 1. Rank my posts by views. Then separately by shares and saves (those predict growth better than views). 2. For my top 10: what do they have in common? Look at topic, hook style, format (talking head / screen recording / text overlay), length, and how each one opens. 3. For my bottom 10: what do THEY have in common? Tell me what to stop making. 4. Find my "outliers" — posts that did at least 3x my median views. For each one, explain the repeatable format underneath it in one sentence, as a recipe I could film again with a different topic. 5. Finish with a list: "MAKE MORE OF THIS" (5 specific video ideas based on my proven winners) and "STOP MAKING THIS" (the patterns that consistently underperform). Be blunt. Use my real numbers in your answer so I can check your reasoning.

Why this beats every "content ideas" prompt you've tried

Generic prompts give you generic ideas because Claude has nothing to anchor on. This prompt forces every recommendation to trace back to a real post with real numbers on YOUR account. When Claude says "make more screen-recording demos", it's because your screen-recording demos are outperforming — not because that's the vibe on the internet this week.

Feed it your competitors — connect Apify

Your own data tells you what worked yesterday. Your competitors' data tells you what's working right now. Apify is a scraping platform with a ready-made Instagram scraper — you give it a list of accounts, it returns their recent posts with view counts. No coding, and the free monthly credits cover a weekly scrape of five accounts.

1

Pick 5 creators to watch

Not the mega-accounts — pick creators one or two steps ahead of you in YOUR niche, whose audience looks like the audience you want. Five is plenty; more just adds noise.

2

Set up the scraper

apify.com → search "Instagram scraper" → paste your 5 profile URLs → run. It returns each account's recent posts as a spreadsheet-friendly file. Ask Claude to walk you through it the first time — it takes ten minutes.

3

Give the results to Claude with this prompt

This is the outlier hunt — the maths that separates "this creator is popular" from "THIS specific post broke out and here's the format to borrow".

Here's a scrape of recent posts from 5 creators in my niche. Find the OUTLIERS — the posts that broke out, not the posts that are merely fine. 1. For each creator, work out their median views across the scrape. A post only counts as an outlier if it did at least 3x that creator's own median. (This matters: 100k views is normal for a big account and a breakout for a small one. Always compare each creator to themselves.) 2. List every outlier: creator, views, ratio vs their median, date, and the link. 3. For each outlier, break down the FORMAT underneath it: the hook style, the structure beat by beat, why a scroller stops on it. Describe it as a recipe, not a summary. 4. Flag any format that shows up as an outlier for MORE THAN ONE creator — that's not luck, that's a wave I should catch this week. 5. End with the top 5 outliers I should recreate, ranked, each with a one-line version of how I'd film it for my audience: [describe your audience here]. Do not tell me to copy anyone's content. I want their formats — the repeatable structure — with my topics.

The 3x rule is the whole trick

Don't chase raw view counts — chase the ratio. A post doing 3x its creator's normal performance means the FORMAT did the work, not the account size. Those are the only posts worth reverse-engineering, and most people never do this maths.

The Monday loop — one sitting a week

The two prompts above are powerful once. The compounding happens when they run every week — because what's working shifts constantly, and the system keeps you pointed at this week's answer instead of last month's.

My actual Monday, start to finish

1. Refresh the data

Re-run the Apify scrape, pull the fresh Windsor numbers. With Claude Code this is one command; by hand it's ten minutes.

2. Claude builds the film list

The film-list prompt below merges both feeds into 5-7 concrete video briefs — hook, beats, format source.

3. Film the batch in one sitting

Read the hook, ad-lib the body, next. A week of content in an afternoon because zero decisions happen on camera.

4. Schedule and walk away

Load the batch into your scheduler, spread across the week. The only daily task left is replying to comments and DMs.

Total: one focused sitting. Not "be on social media all day" — one sitting, backed by data, then back to running your business.

It's filming day. Here's my own performance data [attach Windsor export] and this week's competitor scrape [attach Apify results]. Build my film list for the week. 1. From MY data: which of my proven formats is due another run? Give me 2-3 ideas that re-use a format that already worked for me, each with a fresh topic. 2. From the COMPETITOR data: which outlier formats from this week should I adapt? Give me 2-3 ideas, each naming the source post, the format recipe, and my version of it. 3. For every idea, write: the opening line I say to camera (the hook), 3-5 beats for the body, the on-screen text for the first frame, and a one-line reason this deserves a slot ("your demo formats average 2x your median", "this format broke out for two creators this week"). 4. Rank the full list. If an idea is time-sensitive — a trend or news moment that dies by Friday — put it first and tell me. 5. Keep it to 5-7 videos, all filmable in one sitting with just me and my phone. I'll film exactly what you give me, so make every hook something a real person would say out loud.

Level up: let Claude run the whole loop

If you use Claude Code, ask it to build you a small dashboard that does all of this automatically — scrapes on a schedule, finds the outliers, and shows your film list on a local webpage every Monday morning. Mine looks like a mission control screen and I did not write a line of it myself. Describe what you want; Claude builds it.

Make it yours, not theirs

One rule keeps this system ethical and effective: clone the format, never the content. The format is the structure — hook style, pacing, the shape of the reveal. The content is their story, their examples, their voice. Formats are shared physics; content is theirs.

The line, so it's crystal clear

Stealing: re-recording their script with your face.
The system: noticing that "what happens when you ask AI to do [scary thing]" broke out for two creators this week, and filming what happens when YOU ask AI to do something your audience worries about — your example, your result, your take.

And one more habit that quietly does more for reach than anything else: end every reel with a comment keyword. "Comment SYSTEM and I'll send it to you" turns a passive viewer into a DM conversation — the algorithm reads that as high-value engagement, and you capture the person instead of losing them to the scroll.

The two-tier CTA (protect your email list)

1

On the reel: keyword → a free resource that collects their email. Cold viewers won't buy from a 30-second reel, but they'll happily swap an email for something genuinely useful.

2

In the emails after: that's where your paid offer lives. You own the relationship now — no algorithm between you and them. Never pitch the paid thing directly on a cold reel; you lose the email AND the sale.

Setup checklist — 20 minutes, start to finish

A Claude account (Pro is plenty)

Windsor AI free trial, accounts connected

Apify free account + Instagram scraper

Your list of 5 creators to watch

Any coding ability

A developer, VA or agency

Posting more than you post now

A big budget (see costs below)

📊

Windsor AI

30-day free trial, then from ~US$19/mo. Your own analytics, readable by Claude. This is the "reverse-engineer MY account" feed.

🔍

Apify

Free monthly credits cover a weekly 5-account scrape. Pay-as-you-go after that (a few dollars). This is the competitor feed.

🤖

Claude

The brain. The chat app works fine to start; Claude Code turns the whole loop into one command later. You're probably already paying for it.

📅

Your scheduler

Whatever you already use (Meta's own scheduler is free). The system decides WHAT to film — scheduling stays exactly as it is.

1

Today: connect the feeds

Windsor trial + connect Instagram (5 min). Apify account + first scrape of your 5 creators (10 min). Save both exports somewhere you'll find them.

2

Today: run prompts 1 and 2

The Windsor prompt on your data, the Apify prompt on the scrape. Read what comes back — this is usually the moment the penny drops about what your account actually rewards.

3

Next Monday: run the loop for real

Fresh scrape, film-list prompt, film the batch, schedule the week. That's the whole system. Repeat every Monday and let the compounding do its thing.

The honest caveat

Claude decides what to film — you still have to film it. This system removes the guessing, the blank page and the "what's everyone else doing" spiral. It doesn't remove pressing record. That part stays gloriously, annoyingly yours.

Give Claude control of your socials

Give Claude control of your social media

What's inside