Are you getting tired of ChatGPT’s endless regurgitation of SEO-optimized articles and social media posts? Even ChatGPT is getting tired of that.
Too many people use the trendy AI as a text generator.
Too few use it as a highly capable writing partner that never sleeps.
When I first ran into ChatGPT, I swore I’d never use its output verbatim. Between it being easily detected and occasionally completely fabricated, it just doesn’t get my style and tone right. But it can certainly help me write better.
The trick? I treat it like an editor, a proofreader, and a disgruntled reader at the same time. I make it an enemy of my text and let it attack it as vigorously as possible. And then, I incorporate that feedback into my next draft.
Here is how I use conversational AI to strengthen my argument by making ChatGPT argue against my writing.
Devil’s Advocate
The most impactful way of using ChatGPT mid-process has been using it as a skeptic. After I have written my first draft, I paste the whole thing into a ChatGPT session and then prompt the AI to act like someone who disagrees vehemently with anything I’m writing about. I’m making the AI into my personal devil’s advocate.
Experience this article as a podcast, a YouTube show, or as a newsletter:
One of the benefits of ChatGPT is that it’s trained on millions of think pieces and opinions. That includes the polar opposites of my personal thoughts and feeling as well. So, instead of having to step outside of my own perspective all the time, I can use ChatGPT to play that part.
I ask the AI to list the top three arguments that someone opposed to my thinking would disagree with and why. The response to this single prompt alone often gives me food for thought that can lead to a complete rewrite of the article. Usually, I am trapped in my own ocean of bias, and just the mention of a counterpoint allows me to burst that bubble. What would an opponent of my school of thought claim? How would they attack even a well-reasoned article?
I then ask ChatGPT to give a few examples of answers to those points of criticism. They’re usually not very good, as they’re generic and unspecific, but they point me in a direction that I can explore with my own arguments.
I run this prompt for every long-form article I write. If the points of contention are strong enough, I immediately try to defuse them right there and then by expanding my article to address these concerns.
This usually doesn’t take more than ten minutes. But it’s time incredibly well-spent.
Arcs and Arguments
ChatGPT is good at a few more things that I occasionally ask it to do with my writing.
Sometimes, I write something that feels unfinished. I then ask ChatGPT about “what is missing from this to make it a cohesive argument,” which tends to highlight the missing piece of the puzzle.
The AI is also surprisingly good at detecting the emotional subtext of any given passage. I often ask, “Which emotional arcs exist in the article, and if they clash” — which regularly highlights a way-too-abrupt change in tone that I can then smooth out.
None of these questions asked of the text are in any way novel or magical, but the speed at which ChatGPT can allow you to reflect on them is the awesome part of this experience.
For Example: Examples
One other area that ChatGPT can help with is the quality and accuracy of examples. In my writing, I highlight real-world business examples and founder journeys. But I write from within my own echo chamber, and what makes a lot of sense to me doesn’t necessarily connect with all my readers. For this, I have found it helpful to ask ChatGPT to “find and explain any confusing examples for an audience of X.” If it finds any such examples, I then ask it to “come up with better examples,” — which tends to surface names and ideas that I missed in my own research and exploration. It broadens both my own mind and the “compatibility surface” of the text.
But there’s one big potential problem here: ChatGPT’s tendency to make things up. Any claim related to the real world should be thoroughly fact-checked outside of ChatGPT. Even frameworks and concepts it suggests need to be researched: they might not exist outside that ChatGPT session. I once asked it to find studies on mental health topics, and the AI gave me ten scientific paper titles, of which only three actually existed.
Fact-Checking
This brings me to another editorial job that ChatGPT can be used for: fact-checking.
Wait, what? Didn’t I just say that ChatGPT hallucinates things all the time?
Yes, which is why I never trust the examples it gives. But that doesn’t mean I can’t treat ChatGPT as a “writing truther.” I ask the AI to “be extremely skeptical and surface the five parts of the article that seem to need to be fact-checked.” I make ChatGPT “look for anything that sounds particularly unbelievable” and “search for misleading phrases.”
Since ChatGPT is pretty much a gaslighting engine at scale —after all, its primary purpose is to come up with believable and convincing text— I have found that it’s pretty good at finding this kind of tension in my written content, too.
I often ask the AI to point out any logical fallacies in my text. It has so far found a lot of confirmation bias, a few No True Scotsmen, and a surprisingly high amount of loaded questions. Unbelievable, right?
Shifting Perspectives
Well, it’s really useful to have a logical reasoning system take a look at my drafts. I want my articles to be inclusive and appeal to not just a tiny niche audience. For that, one final thing I use mid-writing is the perspective shift.
I ask ChatGPT two questions: “From a beginner’s perspective, what is confusing and complicated? And, from an expert’s perspective, what is over-simplified or misleading?” The resulting list of paragraphs and sections gives me ample opportunity to make the article more comprehensible to either audience — if I want. If it’s meant for a specific audience, I skip this step. But any topic that touches the lives of all kinds of people on all sorts of journeys —like whenever I write about mental health— I want to leave no reader behind.
So there you have it. Adversarial writing with generative AI is my way of rounding out an article.
When you use ChatGPT as a part-time writing nemesis, you end up with a piece of writing that is both completely written in your voice and more accessible for readers inside and outside your existing readership.