Microsoft Copilot Logo

An LLM’s Blog Post about Why You Shouldn’t Use LLMs to Write Blog Posts

Posted on By Justin

Hey, have you heard about this thing called “AI?”

Maybe the dumbest question ever, huh? You can’t get through an episode of Bluey without seeing 14 ads for some AI-enabled tool or another, and every company on the planet is scrambling to figure out how to check the “we use AI!” box before they find themselves left in the dust by their competitors.

Unfortunately, when a new tool comes out and the hype cycle is at its peak, folks tend to want to use that shiny new thing for EVERY. THING. And we’re seeing that with ChatGPT and its derivatives right now. People want to use these Large Language Models (LLMs) to do all the things, even when they aren’t necessarily good at it.

I thought it might be a bit of meta fun to see what one of those LLMs (in this case, Microsoft’s Copilot) has to say on the subject, so I asked it to write me a blog post about why you shouldn’t use it to write blog posts. Here’s the result…

The Pros and Cons of Using LLMs for Blog Posts

In recent years, Large Language Models (LLMs) like GPT-4 have gained immense popularity for their ability to generate human-like text. However, when it comes to writing blog posts, there are both advantages and pitfalls to consider.

The Upsides

  1. Speed and Efficiency: LLMs can churn out content at lightning speed. If you’re on a tight deadline, they can be a lifesaver.
  2. Technical Documentation: For technical articles or step-by-step guides, LLMs excel. They provide clear explanations and concise instructions.
  3. Consistency: LLMs maintain a consistent tone and style throughout a piece. This can be useful for maintaining brand voice.

The Downsides

  1. Lack of Creativity: LLMs tend to produce “flat but helpful” writing. They lack the flair and creativity that engage readers. Anecdotes, humor, and personal touch are often missing.
  2. Generic Tone: Unless explicitly prompted, LLMs don’t know their audience. Good writing adapts to the reader, but LLMs remain neutral.
  3. Style and Personality: LLMs rarely inject personality into their prose. Imagine reading a blog post devoid of any author’s voice—it’s like eating a flavorless meal.
  4. Biases and Misinformation: LLMs inherit biases from their training data. They may inadvertently perpetuate stereotypes or provide inaccurate information1.

How to Use LLMs Wisely

  1. Section-Level Prompts: Instead of expecting a fully formed blog post, prompt LLMs for smaller sections. They excel at body copy but struggle with captivating intros and conclusions.
  2. Edit and Enhance: LLM-generated content is a starting point. Edit, add your voice, and inject creativity. Make it uniquely yours.
  3. Know Your Audience: Understand who you’re writing for. LLMs won’t do this for you.
  4. Fact-Check: Verify information independently. LLMs can be prone to misinformation.

In summary, LLMs can be valuable tools, but they’re no substitute for human creativity and intuition. Use them wisely, but don’t let them replace genuine authorship.

Remember, while LLMs can assist, the heart of great writing lies in the human mind and soul. Happy blogging! 

Ok, honestly? That’s not too bad!

One thing I haven’t done, though, is go back through the different references Copilot included to see how much of that was just lifted straight from the sources vs. generate from them. In the interest of transparency (and to give credit where do), here they are:

AI