I have been using AI to write better content faster for almost two years. In that journey, I enjoyed many advantages like saved time and reduced cognitive fatigue. However, there were many days when wrestling with LLMs was a bad idea.
This article dives into two pros and two cons (broad level) from my experiences with many AI chatbots, especially ChatGPT and Claude 3.5.
Pro: Expand and validate
As you know, LLMs predict text based on the pattern they recognize in your text and in their database. Consequently, they are amazing at expanding your ideas which allows you to visualize your potential content piece tangibly.
You can fail fast and iterate quickly to discover the best angle for the topic you have in mind which increases the overall efficiency of your efforts.
Back in the day, I had to scrap many drafts midway after realizing they wouldn’t have the desired effect on the reader. Fortunately, I can quickly generate short test drafts of titles for specific audiences to get a taste of my ideas.
There are a couple of ways you can do this:
Directly generate the draft. Provide details such as the word count and target audience. Try not to provide too many instructions about the articles. You are more likely to get a better result.
Multi-perspective critique of the idea and suggestions to improve upon it. Share the title and what you wish to discuss in the article. The LLM will then present you with multiple avenues of thought that can help you enrich your piece.
The ability to quickly expand and validate your content ideas allows you to choose the most effective titles for your blog, website, or any other platform. You can visualize how your article will be received by your audience and proceed accordingly.
Con: Increases dependability subconsciously
After the first few weeks of leveraging AI to enhance my content writing process, I was elated. It almost felt like a cheat code to life and it still does. Since I am a human, I occasionally got swayed and used it in instances I shouldn’t have, which backfired massively.
People who write to make money produce two kinds of articles — for the search engines and for the audience. Think of it as a spectrum, some articles are more SEO-leaning than others.
Writing for SEO is pretty straightforward. You target a few keywords, write topics around it, and build links to make it look more authoritative in the eyes of Google. As it can be algorithmized more easily and effectively, ChatGPT can write them well.
Even today, when I am writing a piece for purely SEO purposes, many GPT-generated sentences and even some paragraphs can be directly copied to the final draft. They come out as coherent sentences with organic usage of the relevant keywords.
However, there are plenty of articles that aren’t SEO-leaning at all. This could include messages from the C-suite executives, an announcement from the company, or a detailed explanation about a product or service.
LLMs suck balls at generating these.
And, as an engineer, I knew that. Unfortunately, I was enticed by the sweet gains of time savings and decreased cognitive loads. After the drafts were generated, I would curse my fate and rewrite the whole piece from scratch.
Sometimes I even caught myself creating article outlines that would yield better drafts with ChatGPT. This was huge considering AI tools should be a part of the content writing process, not the determinants of it.
Pro: Automate boring stuff
If you think about it, writers only spend half the time actually writing. The other half is spent cycling through tabs, creating files and organizing them, and other administrative tasks. You can mitigate them by using templates and streamlined workflows.
Including ChatGPT or any other LLM in your content production process will expedite a few of these boring yet essential action items. A common way you can use them right now is by leveraging them to write meta details.
The Article Meta GPT takes in an entire article draft to return the following:
10 meta details
10 alternate title suggestions
5 URL slug suggestions
3 LinkedIn posts
You can also drop the published link of the article or upload a PDF directly. But be mindful of the context window limitations of LLMs.
Another productivity tool is the Quick Answers GPT. Sometimes we need short answers to certain questions that pop up in our heads while writing an article. “Wonder what this means?” or “Are there any better synonyms for this word” types of questions might be common.
Typically, GPT returns an essay-style answer in a bid to be thorough and comprehensive which we don’t have the energy or fucks for. When writing for clients, you just need to understand the terminologies just enough to push out a persuasive piece.
The Quick Answers GPT does just that. You will get your answers within 50 words. It is also programmed to search the internet on time-sensitive topics like weather and sports.
I am aware that the range and scope of such administrative tasks vary from workflow to workflow as they depend on tens of factors, a lot of which are intangible. So, in a later chapter, I have revealed a thought-process that will help you conceptualize, test, and implement custom GPTs for your use case. So, stay tuned.
Con: Potential time waster
For every 10 hours you spend on an LLM, you spend one that leads to nothing. It could be writing trash prompts because you are feeling lucky or making sense of AI’s responses that are just hallucinations.
And that’s when you already have an established workflow with the right tools and have tried and tested prompt templates. When you are experimenting with LLMs to build the said workflow, the loss could be as much as 100% on certain days.
I reckon I have written thousands of words in prompts that had to be dropped after a couple of days of usage because they were just shit. There are also a handful of GPTs that I’ve wasted significant resources on which are now collecting dust.
A simple workaround is to use my established ChatGPT content-writing workflow that enables users to significantly personalize their prompts. However, as awesome and versatile as that workflow is, it can still fail, especially when you are writing extremely niche topics for a narrow audience.
With all of that being said, I think the wasted time is still worth it. If you stick along long enough the gains are massive in terms of productivity and output as a writer. Just keep learning and improving with each passing draft and the benefits will arrive.
Conclusion: It all depends
Leveraging AI for content creation or any type of serious work demands a measured and calculated approach. The applications of these LLMs are so vast that without a weighing scale, you will be lost.
Start simple. Pick one task that you wish to automate partially or wholly with AI. Write a prompt, test and iterate, and implement. Rinse and repeat.
Putting your hands in lots of cookie jars is unproductive. Take it from my mistakes. Looking back, a lot of those hours wouldn’t have gone to waste if I just built linear processes around testing prompts. But that’s the game, I guess.
—
Next Chapter: 5 Ways LLMs Enhance the Content Writing Workflow
Previous Chapter: Best Practices While Using ChatGPT to Write Content
Index (with Prologue): Content Writing With AI: Principles, Strategies, and Prompts
Comments