#7: How I use 2 Custom GPTs to write viral LinkedIn posts (Tutorial)
Created with Canva

#7: How I use 2 Custom GPTs to write viral LinkedIn posts (Tutorial)

Generative AI can write very successful posts. But you're probably using it wrong. And then you blame the tool. That's not fair. How are you going to explain that to the robot overlords, once the machines take over?

In this edition of Humanly Analytic, I am going to show you how I used several Custom GPTs to create that post. At the end, you will be able to create those tools for you and up your content, without becoming a generic post robot.

But let's look at the facts first. My most viral post in 2023 reached the following numbers:

?? 6414 likes.

?? Over 370,000 impressions.

?? 483 comments.

?? Featured by LinkedIn News.

And I (only) have 15.400 Followers.

How I teach ChatGPT to write like me

Many say that #ChatGPT only writes generic and impersonal content. Yet, I think it's because many people are not getting the most out of these kinds of tools. They type a short prompt and expect the machine to do the magic.

The biggest strength of ChatGPT is not to create FOR you. It's the ability to co-create with you.

First, you will need a CustomGPT that turns your basic idea into a solid post. If you cannot see the "Create GPT"-option (screenshot below), you'll probably need a plus license to enable it.

Start creating a GPT by clicking the "Create"-button of ChatGPT

This CustomGPT needs to be prompted, so that it interviews you. It gathered all information that's floating around in your head. It also asks you about your audience and desired tonality. This can easily be done in the editor window of your CustomGPT (screenshot below). The cool thing: You don't need to code.

Designing your GPT by simply describing it in natural language

Now, it should also learn from (some of) your past posts, so that it creates a draft that resembles your style of writing. In order to let your CustomGPT learn your style, you can use your official LinkedIn post statistics by following these steps.

  1. You need to download you personal LinkedIn dataset, which you can officially request in the settings & privacy area of your LinkedIn settings (screenshot below)
  2. You need to anonymize the dataset and only keep the data about your posts' text, engagement stats and posting dates
  3. You need to upload this anonymized dataset to the knowledge base of your CustomGPT

Click on "Get a copy of your data" in the LinkedIn settings menu.

Congratulations. Your GPT now draws from the statistics about your past LinkedIn posts and is also able to build on your writing style. It also has knowledge about which kind of post was more or less successful. Which topics gain more attention. This is super helpful for creating authentic and personalized content.

How I create LinkedIn post with the help of ChatGPT

The magic happens, when you have managed to instruct your CustomGPT in such a way, that it starts a reflective dialogue with you. You may tell it a basic idea, which you want to write about.

My GPT named "Astrid Grindlen" specializes in creating story-posts

Example: I want to write about the importance of data & tech literacy in today's workplace.

Based on the way I instructed my GPT, it will gather information about why this topic matters to me, how I am affected by it, and how I am planning to solve it.

The conversation is the important part.

It will ask me about my key learnings regarding the topic. It will also ask me, which kind of post I want to create (e.g., a "text-only-post" or an article).

This dialogue is the method by which the GPT is enabled to create personalized posts. It co-creates with me and augments my writing capabilities.

How I bring the post across the finish line using ChatGPT

The best post will not get much attention, if you do not perfect the hook. At the same time, you should not oversell or create clickbait-hooks, as this will turn your audience away over time.

So, you guessed right: I created another GPT called "Hook Hero". Its only job is to ask me for the final version of a LinkedIn post and suggest 3 different hooks. I followed the same steps to create it. However, I gathered the hooks of about 100 viral posts that I also personally liked for their style.

I uploaded this data to the GPT, so that it can use these examples when suggesting hooks for my posts.

Why not just write the post without ChatGPT?

My full-time job takes all most of my attention and I like it that way. I want to be present and engage on LinkedIn, but besides my private life as a father of three and a hopefully okay partner in crime, I just don't have the time for manual posting. And I don't want to make more time for it.

But I also do not want to create generic posts. My community has grown, because I am trying to be thoughtful about my content. At least that's what my peers and colleagues tell me.

So, there is no magic bullet. You'll still have to invest some time into crafting great content. But, I would estimate to be about 10-20% more efficient using my GPTs, while not seeing a dramatic negative impact on the reach and engagement below my posts.

And, to be honest (because David Beckham wants me to): I only create some of my posts with these GPTs. You can tell, because I include a footnote in them. Did you spot it already?


Liked this piece? Keep on separating fiction from facts with my German newsletter: HRDL News or my English Newsletter: Humanly Analytic.


This article reflects my personal views only and is not necessarily the view of the companies I am associated with.

That's an interesting approach, Daniel. And the hint to download the data should also give us more possibilities to analyse them. However, after downloading my data, I'm a bit lost. I see my posts in the shares.csv but I don't see the posts' performance. Neither views nor reactions. This data would be indispensable as DV for any algo to create performing posts, I presume. Am I missing something in the files or did you combine the data with the performance of the best 50 posts that can be downloaded from the analytics page? This seems complicated as the links for one post in both files are not the same!

要查看或添加评论,请登录

社区洞察

其他会员也浏览了