Course Unlocked
Posts
The Best AI Tool for Social Media Growth is Not on YouTube

The Best AI Tool for Social Media Growth is Not on YouTube

The best AI tool for social media doesn’t exist. So I built a system instead.

Jonathan Delamater
April 10, 2025

DISCLOSURE: This post contains affiliate links, which means I may earn a small commission if you buy—at no extra cost to you, sometimes you might even get a dealio, you never know! But, I only share tools I believe in, because your trust means much more to me than any quick buck. 👊 Cheers!

Course Unlocked

Growing a social media audience is hard friggin’ work.

Not “clock in, clock out” hard.

More like “Sheesh, I know I want this - but after work, I just want to chill and crack a beezy…” hard.

I’m convinced Hormozi is a machine - I’m not missing White Lotus, ok.

And no, I’m not about to hand you the same gluten-free advice being regurgitated across every platform that still has a pulse.

📈 “Post consistently.”
📈 “Engage more.”
📈 “Provide value.”

Sure. While we’re at it, I’ll just add “work harder” and “do better” to today’s to-do list.

You’re welcome.

👀 What This Actually Is

This is me, naked and afraid at the keyboard - doing what’s known as “building in public.”

Not mid-launch, not post-exit, just a dude with a job, a family, limited time, and an allergy to B.S. content that is the marketing equivalent of meatloaf wearing a bowtie.

But… what I’ve been building is kind of working. And slowly, I’m making it even better - I thought it was worth sharing!

So if you’ve got seven minutes and a set of eyeballs, let’s friggin’ go!

Again, most AI-generated content sounds like a motivational poster got passed through a woodchipper.

I myself, tried automating my posts using the many techniques “as seen on YouTube” that are floating around out there in the wilds of 2025.

Spoiler: It’s hot garbage.

It all looks so clean in the tutorial, right? “Just copy this Zap, connect it to your Notion database, and boom - viral content forever!”

Except… no. No boom. Just a quiet thud and the sound of your engagement crawling into a corner to die.

Because here’s the thing they don’t tell you in those 14-minute how-to videos:

Their processes are just way too simple to be effective.

This is one of the most complex ones I could find:

A Screenshot of a YouTube Video About Posting To X On Autopilt Using Mak.com - Simple Workflow

69K Views - 1 GPT Module

I’m not trying to single out this creator either - this is one of the more complex iterations. But, I can tell you…

There are so many friggin’ problems with these types systems!

Conveniently, none of which are mentioned until you're too deep to back out.

So I thought I could help save you the headache!

😬 Why AI Content Flops (Even If It’s Technically Fine):

It has NO CONTEXT.

These were the core issues I found:

Same source material as everyone else. Just regurgitated ideas from the internet’s cesspool - zero originality, max déjà vu.
No memory of past posts. It couldn’t remember what I’d already said, so it kept pitching ideas I’d already written, deleted, and emotionally processed.
Bad Content Ideas. Using ChatGPT to create new, relevant, even interesting content ideas is a task and a half. And saving ideas in a Google Sheet or using an RSS feed just requires work I don’t want to think about, let alone do. I don’t want to recycle news posts, I want to make valuable content. But, that’s just me.
No Data. How will it get better at posting good content for me over time? Answer: YouTube hasn’t figured that part out yet.
No sense of humor. I have AI help me write these blogs and boy do I struggle.

Without this context. It’s just empty.

🧠 The Fix: A Stack That Remembers What You Forgot

My Daily X Posting Scenario in Make.com

So I built something better.

Something that doesn’t sound like three robot ducks convincingly suspended in a trench coat.

Here’s what it does:

Uses content ideas from my new and existing content
Knows my brand voice
Stores all my posts internally
Remembers what I’ve already said, so it doesn’t repeat itself
Critiques its own content for quality and humor
Tracks what worked - and what quietly disappeared into the algorithm void

The tools:

Beehiiv – An excellent and feature-rich newsletter and website platform where each newsletter you post can double as a blog post - instant SEO, lead magnet, and website all in one platform.
Make.com – An automation and Integration platform that allows limitless possibilities. Connect your entire tech stack seamlessly.
Pipedream – An automation and Integration platform that allows limitless possibilities. But this tool also allows you to run native Python and NodeJS - Make, take notes here!
Google Sheets – Why bother with Notion or AirTable which limit you to 1000 rows? Get 1,000,000 rows for free by just using Google Sheets as your database.
ChatGPT (Assistants API) – For creating your post content and ingesting your data, brand voice, and more.

Important Note About Posting To X Via Make.com

Make.com has deprecated their native X integration module.

After struggling to find a workaround for days - messing with Oauth1, Oauth2, V1.1, etc…

I’ll spare you the explanation and just give you the solution!

Create your post text using Make.com (as shown in the workflows around this section) and follow all other steps as always - but to post the actual tweet text to X, you must:

1.) Set Up a Free Pipedream Account

A.) Create a new project

B.) Set up a new workflow

Click The Blue Button

B.) First Module: Webhook Trigger

C.) Second Module: Node.Js Module

The Actual Pipedream Workflow

CODE:

import { axios } from "@pipedream/platform"
import twitter from "@pipedream/twitter"

export default defineComponent({
  key: "twitter-create-tweet",
  name: "Create Tweet",
  description: "Create a new tweet using the Twitter API v2",
  version: "0.0.1",
  type: "action",
  props: {
    twitter,
    tweet_text: {
      type: "string",
      label: "Tweet Text",
      description: "The text content of your tweet",
      default: "Hello from Pipedream using X API v2! 🚀 #NodeJS #XAPI"
    }
  },
  async run({ steps, $ }) {
    const config = {
      url: "https://api.twitter.com/2/tweets",
      method: "POST",
      data: {
        text: this.tweet_text,
      },
    }

    const headers = this.twitter._getAuthHeader(config)
    
    try {
      const response = await axios($, {
        ...config,
        headers: {
          ...headers,
          "Content-Type": "application/json",
        },
      })

      if (response?.data?.id) {
        return {
          status: "success",
          tweet_id: response.data.id,
          text: response.data.text
        }
      }

      return {
        status: "error",
        response
      }

    } catch (err) {
      return {
        status: "exception",
        message: err.message
      }
    }
  }
})

D.) Deploy the workflow for production use.

2.) [In Make.com] Use a HTTP (“Make a Request”) Module at the stage where you want post the actual tweet and you have already generated the text for that.

Settings:

URL → Your Pipedream Webhook URL (Check the module in Pipedream to find this)

Method → select: “Post” request.

Headers → (Add a header - it should then display two boxes)

First Box (Name): Content-Type

Second Box (Value): application/json

Body → Select: “JSON.”

{ “tweet”: “{your tweet text from a previous module}” }

Make.com Scenario - Beehiiv to X Workflow

This Catches My Posts And Generates Social Content / Updates The Database

🤯 How It Works (In Human Terms)

I write my actual thoughts - not AI drafts - as blog posts/newsletters (like this one) in Beehiiv. Though, I do run my initial brain dump through AI afterwards for better structuring, brand voice consistency, etc. Then I edit again… and again. Finally, I schedule the post, and usually, on Fridays, when the post drops a webhook I set up in Beehiiv sends my post contents directly to Make.com.
Make.com catches the post once it drops, and stores it in a spreadsheet. It also stores it in my ChatGPT Assistant.
The sheet logs it with a few other key items though:
- Content
- Date
- Topic
- Tags
- Ways to differentiate from this content
- Meta Description
- Hook Summary
- All Email/Web Metrics - though, if the blog just posted it will probably have none, I just mention this because there is another scenario I have which updates these over time, so they do populate.
GPT reads the sheet to look at other recent blog and social media posts - but through the OpenAI Assistants platform, it also has prebuilt context as to:
- My brand voice doc
- My brand guidelines
- My blog post archive
- A running list of content I’ve already posted to social platforms (separately)
It generates an X post based on the new blog post content (Though I am reconstituting these scenarios to also work for Facebook, and LinkedIn, and possibly others)

So it’s like having a GPT, trained on my brand and content - but it’s usable outside the ChatGPT interface. It’s not perfect. But it’s not dumb, either.

And honestly, that’s already better than half of what’s on the explore page.

🧪 Why X Is the Best Way To Go For This:

Sadly, you can’t post junk on LinkedIn. Well - you can post junk… but then you risk your entire profile getting ranked as “mediocre” by the algorithm. A fate worse than marketing hell.

Instagram? You breathe wrong and your engagement drops 30%.

TikTok wants you to lip-sync your trauma for views.

But X… X is chaos with a character limit.

You can post 10 times a day and no one blinks an eyelash - It’s a testing lab in disguise.

So I post. Track. Repeat. (I’m doing 3 posts a day right now - unless I post a blog that day - then, 4).

What works and gets engagement becomes future content - likely a short or a long form video. What flops becomes a footnote in my self-awareness.

Screenshot of a Make.com workflow that takes new beehiiv blog posts and uploads them to my GPT

This Scenario Keeps My AI Assistant Trained On New Content and Updates Existing Content

🔁 The Feedback Loop That Keeps Me From Starting Over Every Week

AI without feedback is like yelling in an empty room and calling it a strategy.

So I log every AI-assisted post.

I track:

Like Count
Share Count
Comment Count
Save Count (depending on the platform)
Impressions (depending on the platform)

Over time, some of this content (depending on the API) will be automatically updated in other Make.com scenarios - in my Google Sheet that stores each post’s metrics. Some (like X) I have to update manually - but think about it - 20 or so posts a week takes me less than 15 minutes to update weekly - whatever.

GPT learns. I learn.

And eventually, the system starts writing like someone who’s been paying attention.

🎯 So… What’s the Best AI Tool for Social Media?

It’s not a tool.

It’s the system that makes decisions for you when you’re too tired to pretend you care about engagement rates. It learns, creates original, compelling, and quality content.

So, again - here’s my setup:

Beehiiv – The source of truth.
Make.com – The invisible hand.
Pipedream – The bridge to X.
Google Sheets – The brain.
ChatGPT (Assistants API) – The ghostwriter.

It’s not flashy.
It’s simply built to last, hoss.

🧠 Want me to share how I did all of this? Just say the word! (Reply to this email or message me on X)

👉 Subscribe to the newsletter
👉 Follow me on X
👉 Or just sit quietly and let the ideas come. That works too.

Until next time, sayonara!

Reply

or to participate.