Disclosure: this article was written by a human.

It seems like everyone is trying to sell you on the one AI tool that is going to solve all of your problems. It feels sometimes like we’ve fully reached Peak Hype in the Age of AI™.

But, cynicism aside, ChatGPT and the like are still useful, but more so when used to supplement a human than when just left on their own.

Before we get into the useful ways to use LLMs, let’s look at the wrong way to use an LLM.

Bad uses of AI for writing a post

This is probably not the way you’re writing blog posts, but admit it, you’ve tried it 😂

“Write me a blog post about XYZ…”

And it was pretty bad, wasn’t it?

Why it doesn’t work

No matter how much prompting, instruction, rules, or details you provide to the model, the output is still only plausibly good. The content will still:

• feel generic,

• sound robotic,

• all the paragraphs will be approximately the same length,

• each section of the post ends with a weird summary,

• and the article overall will not really say anything meaningful or inspiring.

I even asked ChatGPT why it’s writing is so bad, and it told me:

Creating an engaging blog post requires a blend of several elements, which can sometimes be missed by AI-generated content. Here are a few reasons why an AI-generated blog post might feel bland:

• Lack of Personal Touch: Missing unique voice or personal anecdotes.

• Limited Creativity: Sticking too closely to generic templates.

• Depth and Insight: Lacking deep insights or nuanced perspectives.

• Emotional Connection: Not effectively conveying emotions or connecting with readers.

• Storytelling: Struggling with crafting compelling narratives or story arcs.

• Engagement and Interaction: Missing interactive elements or calls to action.

You can definitely ask ChatGPT how to improve it’s writing style, and it will tell you, but I still find the returns still diminish over time.

When asked about it, ChatGPT even admitted:

Using LLMs as tools to assist and augment human creativity can produce more innovative and high-quality results.

This is because LLMs operate on plausibility. They are just trying to guess the next word in the sequence based on how probable it is. What that really means is that it will always be a little bit wrong.

So then, what’s a better way to use AI for our conceptual blog post?

Good uses of AI for writing a blog post

I tend to look at tools like ChatGPT as a way to supplement my current writing. Something that ChatGPT is good at is evaluating something that already exists, and comparing it to a set of expectations.

Here are some examples.

Generate some blog post ideas:

Act as a world-class marketing expert. I am in the field of XYZ and interested in developing some blog posts about latest trends within my industry. Help me come up with 5 ideas that might make an engaging blog post.

Evaluate an existing draft and make it better.

Act as a world-class copywriter, evaluate the attached blog post and make recommendations about how it might flow better, if there are any consistency issues, or missing pieces.

Rewrite a post in a XYZ voice:

Act as a world-class neuro-linguistic analyzer & copywriter, evaluate the attached blog post and make recommendations on how to make it sound more [formal|funny|serious|person XYZ].

Shorten an article:

Act as a world-class copywriter. Shorten this article to fit within 1500 words, without changing the author’s voice.

Get relevant details from a related article:

Visit this article at https://some-article-url.com/xyz-article and give me a brief list of key takeaways from the article that are relevant to the attached blog post.

Generate a good blog post summary

Analyze the attached blog post and generate a summary that is SEO friendly that is less than 40 words.

Generate a social media post about the article:

Act as a social media & copywriting expert. Analyze the attached blog post and generate a social media post appropriate for [Facebook|LinkedIn|Instagram|etc]. Include [emojis|humor|hashtags|etc] in the output and keep the output within X words.

Generate image search terms that suit the mood of the post:

Act as an image search specialist. Generate a comma-separated list of keywords that could be used to find a stock image that would be appropriate to accompany this blog post

In this last example, I almost always use paid stock images because as much as I enjoy AI, I still value supporting actual artists.

Wrapping up

I hope that gives you some good ideas, and helps you get more out of your sessions with ChatGPT.