Everyone’s using AI for their writing.
In fact, I collaborated with an AI to write the piece you’re reading right now. But I bet you won’t be able to tell which parts were mine and which belong to ChatGPT- and that’s the key to an effective human/AI writing collaboration.
AI is going to make writers more scalable, but not extinct. We’re moving to an editing role, and will likely still be on the hook to write the most important parts of a piece- like the introduction, for example- and make sure everything sounds natural.
If you’re not a great writer, it’s a headache to refine a piece that is already ‘pretty good.’ That’s why companies are still going to need great writers, and their content is still going to be a notch above everyone else.
But don’t worry- you don’t have to be a great writer to take some basic steps with AI-generated content that will help you pass it off as copywritten.
The Uncanny Valley
Have you ever seen a computer-generated image of a human that just looked… off? Like something was just a bit strange or unsettling, even if you couldn’t quite put your finger on it? That’s what we call the uncanny valley.
‘Uncanny valley’ is that funny feeling we get when we encounter something that’s almost human, but not quite. The concept was first introduced in the 1970s by a roboticist named Masahiro Mori. He observed that as robots became more human-like in appearance, there was a point where they became too human-like and caused a sense of revulsion in humans.
Just like a weirdly simulated 3D character, AI-generated written content can sometimes come off as artificial or robotic. Sure, the words and sentences might be technically correct, but they don’t have the human warmth that makes for engaging and relatable content.
If you want to use AI content, you have to take steps to bridge the uncanny valley. By injecting some humanity and personality into the words, you can create content that actually connects with your readers, without triggering the feeling that they’re about to be turned into organic robot fuel.
How to make AI-generated content sound more human
1. Use the right prompts
Interestingly, this was one of the few concepts ChatGPT couldn’t quite wrap its head around. And it makes sense- the language “AI prompt” hadn’t entered the public discourse until after ChatGPT had already been trained.
But that’s changing very quickly. Job postings for “Prompt Engineers” are popping up on career sites, offering impressive salaries. It’s clear that knowing what to say to get exactly what you want from AI is going to be a skillset all its own.
Here’s what I’ve found works best:
- Be specific about your audience, purpose, and platform. Tell the AI exactly what you need, where you’ll be sharing it, why you’re writing it, and who will be reading it.
- Describe the style you want. Use words like “conversational” or “relatable language.” Ask for “dry humor” or “a touch of sass.”
- Use examples and analogies. For example, you could say, “rewrite this like you’re explaining it to a friend who knows nothing about mental health,” or “make it sound like you’re telling a story at a dinner party.” A personal favorite of mine is, “write this like you’re a stand-up comic.”
2. Define your desired tone and voice
The cool thing about ChatGPT is that it can learn and reference things you’ve discussed, but only if those are contained earlier in that same conversation.
That’s why it’s important to keep everything in one conversation- start that new conversation off by talking about your style, tone, voice, and brand “personality.” Then, when you need ChatGPT to write something for you, you can return to that conversation and reference the tone and style you discussed.
But here’s the thing- you need to let ChatGPT find the words that define what you’re looking for. For example, what you call “empathetic,” ChatGPT might describe as “tender.” And what you think is “fun,” ChatGPT might see as “erratic.”
That’s why you should feed ChatGPT sample pieces and ask it how it would define the style and tone. Then, you can use those terms in the future when asking for content.
Any good copywriter can tell you: less is more. The mark of a good writer is the ability to use fewer words, more powerfully.
And if you’re looking at your meticulously generated AI-generated content, I’ll bet you’re noticing that ChatGPT is a bit of a show-off.
It will constantly tend towards bigger words, longer sentences, and blocks of text that make it hard for readers to sail through what you’re trying to say. It also has this terrible habit of adding greetings and transition words in awkward places.
So your first line of action is to simplify the text.
Read through and tweak anything that’s too long, too complex, or too high-level. Break up run-on sentences, shorten paragraphs, and add spacing between ideas. Remove jargon. Anything you can’t see yourself saying in conversation, take out of the piece.
4. Add humanity
There are certain things that can only come from a human, like true stories that connect with the heart of the issue, incredible metaphors, and emotions.
I would argue that you should never outsource anecdotes to ChatGPT- not only because AI isn’t very good at it, but because it’s unethical to lie to your readers.
I also tend to write every introduction and conclusion. These are key parts of a piece of writing that help anchor a reader to the voice of the person speaking- the tone matters most there.
But there are other things that you should drop in to take that AI content to the next level. For example, curse words- if it fits your brand vibe. Contractions if it seems too formal. Jokes. Cultural references. Non-sequiturs. Anything that can add some human “quirk” to the robot output.
5. Fact check
Chat GPT is the most bold-face liar I’ve ever seen.
And it is as smooth as a courtroom lawyer.
That means you need to fact-check your AI-generated content within an inch of its life.
Names, dates, facts, research studies- don’t trust anything.
Even if you think something is mostly true, double-check- sometimes, it fudges dates or slightly misspells names.
There’s no quicker way to undermine your credibility than to fudge up simple facts by delegating information gathering to an unreliable AI.
You can outsource writing, but you can’t outsource being human.
So here’s a conclusion written completely by me (because ChatGPT always starts conclusions with “in conclusion,” which I haven’t done since high school English class).
I’ve already heard from multiple companies that they’ve stopped hiring copywriters- which makes sense.
But they shouldn’t bench all their writers. The uncanny valley effect is real, and writers will know how to take AI content from robotic to relatable.
As I prepare to take on my first few AI content editing projects, I really believe we writers can bridge the gap between AI and authenticity. And it’s a worthy goal to make sure that excellent writing retains its uniquely human warmth and charm.
Simplify language: Replace complex or technical terms with simpler, more accessible language.
Vary sentence length: Use a mix of short and long sentences to create a natural flow.
Use contractions: Contractions make writing sound more conversational and less formal.
Add personality: Inject your brand’s personality and tone of voice into the content to make it more relatable and engaging.
Remove jargon: Avoid using industry-specific jargon that could confuse readers.
Break up large blocks of text: Use subheadings, bullet points, and short paragraphs to make the content easier to digest.
Fact Check: Make sure AI didn’t botch names, dates, facts, or references.
Use examples: Incorporate examples and stories to illustrate key points and make the content more relatable.
Read it out loud: Read the content out loud to make sure it flows naturally and sounds like something a human would say.
Did I miss anything? Let me know!
Got some AI content that needs some personality? Email me at email@example.com.